An Orally Active Galectin-3 Antagonist Inhibits Lung Adenocarcinoma Growth and Augments Response to PD-L1 Blockade

Globally, lung cancer is the leading cause of cancer-related mortality (1). Non–small cell lung carcinoma (NSCLC) comprises 80% of total lung cancer cases, with lung adenocarcinoma being the major subtype (1). In recent years immune checkpoint therapies targeting various negative regulatory receptors on tumor infiltrating cytotoxic T lymphocytes (CTL) such as programmed death-1 (PD-1), programmed death-ligand 1 (PD-L1), cytotoxic T-lymphocyte associated protein 4 (CTLA-4), and others, have shown unprecedented efficacy in NSCLC patients even against late-stage disease (2). However, patient response is limited, thus driving intensive research toward combining immune checkpoint inhibition with other targeted agents to overcome resistance (2).

Tumor-associated macrophages (TAM) are present in the stroma of many tumors including NSCLC (3). TAMs acquire an alternative (M2)-like macrophage phenotype and secrete angiogenic and anti-inflammatory cytokines, which contribute to the immunosuppressive milieu of the tumor microenvironment (4). TAMs can also be important direct targets of PD-1/PD-L1 inhibition and can also promote drug resistance by removing anti-PD-1 antibodies from T cells (5, 6). Indeed, macrophage depletion via colony-stimulating factor-1 receptor (CSF-1R) blockade improved T-cell infiltration and antitumor activity of PD-1 antagonists in preclinical models of melanoma and breast cancer (7, 8), suggesting that strategies aimed at inhibiting macrophage responses are necessary to permit effective immune checkpoint therapy.

One possible target for such combination treatment is galectin-3, a member of a protein family defined by affinity for β-galactoside-containing glycoconjugates and a conserved carbohydrate-recognition-binding domain (9). Galectin-3 is widely expressed in several cell types such as macrophages, fibroblasts, activated T-lymphocytes and epithelial cells (10–12) and is highly expressed in high fatality cancers such as NSCLC (13). In NSCLC particularly in adenocarcinoma, increased galectin-3 expression in tumors, lymph nodes and serum correlates with metastases and is a negative prognostic indicator (13–18). The galectin-3 genetic polymorphism rs4652 associated with impaired galectin-3 secretion, has been linked to increased survival and response to chemotherapy in NSCLC (18). Galectin-3 can directly enhance cell proliferation (19), apoptosis resistance (20), metastatic potential (19, 21), as well as lung cancer stemness (22). It is also an important constituent of the tumor microenvironment acting on endothelial cells to promote angiogenesis (23). Furthermore many studies have revealed the inhibitory effects of galectin-3 on activated cytotoxic T lymphocytes CTLs (24–27) and we have shown it to be essential for M2 macrophage differentiation (28, 29). Hence, galectin-3 forms an ideal candidate target for combining with checkpoint blockade.

We examined the role of galectin-3 in NSCLC by utilizing the syngeneic mouse Lewis Lung Carcinoma (LLC1) model, comparing tumor growth in wild-type (WT) and galectin-3–deficient mice showing an essential non-redundant tumor-promoting role for galectin-3. Bone marrow (BM) transfer and macrophage depletion experiments show that macrophages are a major source of tumor-promoting galectin-3. A newly developed, selective small molecule galectin-3 inhibitor inhibited mouse and human NSCLC tumor growth and metastasis and significantly potentiated response to an immune checkpoint inhibitor.

LLC1 cells and A549 cells were purchased from the European Cell Culture Collection (ECACC 90020104) and were cultured at 37°C in 5% CO₂ (95% air) in DMEM (Sigma D5671) supplemented with 10% FCS, 1% l-glutamine, and 1% penicillin/streptomycin. Human cell line A549 was authenticated by short tandem repeat (STR) DNA profiling by ECACC. Cells were routinely tested for Mycoplasma every 6 months. Vector pCMV-KDEL-Gluc-1, expressing G.princeps luciferase (Lux Biotechnologies) was transfected by electroporation (Lonza electroporation kit VCO-1001). Stably transfected cells were selected with G418.

All animal experimental work was carried out under a project license approved by the local Animal Welfare and Ethical Review body (AWERB) and issued in accordance with the Animals (Scientific Procedures) Act 1986. C57Bl/6 mice and female CD1 nude mice were purchased from Harlan Laboratories. Generation of galectin-3^−/− mice by gene-targeting technology has been described previously (30). CD11b-DTR (Diphtheria Toxin Receptor) mice were derived from FVB mice as described (31) and backcrossed over 10 generations onto the C57Bl/6 background. Human galectin-3 knockin mice were generated by Cyagen Biosciences using the TurboKnockout (conditional Knockin) approach by inserting the entire human LGALS3 sequence into exon 1 of mouse Lgals3 so that the expression of human galectin-3 is under control of the mouse gene regulatory element.

Mice were anaesthetized with isofluorane. A 1 mm skin incision was made below the right shoulder blade. A total of 103 LLC1-luciferase cells were injected through the intercostal muscles into the lung parenchyma prior to the incision being stapled.

LLC1 (2.5 × 105) cells were injected subcutaneously into the flanks of age-matched male WT and galectin-3^−/− C57Bl/6 mice. Each animal received an injection of 2.5 × 105 cells suspended in 100 μL PBS in both flanks. Tumor volumes were measured with calipers every 1 to 3 days [tumor volume = π/6 × (L × W)^3/2].

LLC1 cells were administered via the tail vein (1 × 10⁶ cells) and lung lobes were harvested at 7 days. RNA was extracted from lung lobes using a Qiagen RNeasy Kit, converted into cDNA (Quantitect cDNA Synthesis Kit; Qiagen), and luciferase expression was measured by qPCR using primers against Gaussia princeps luciferase (5′-TCTGCCTGTCCCACATCAAG-3′ forward and 3′-CCCTGTGCGGACTCTTTGT-5′ reverse; Primer Design) and SYBR Green (ThermoFisher Scientific).

CD-1 nude female mice received 3 × 10⁶ human lung adenocarcinoma cells (A549) in 100 μL 1:1 Matrigel:serum-free DMEM in both flanks. Tumor volumes were measured every 2 to 3 days using digital calipers.

Macrophages were ablated in C57Bl/6 CD11b-DTR mice (or WT littermates) by administration of 10 ng/g diphtheria toxin (DT) intraperitoneally (i.p.) prior to subcutaneous tumor cell injections.

Mice were injected with 400 μL liposomal clodronate (Liposoma). After 36 hours, mice were irradiated with 10.5 Gy delivered from an IBL637 gamma irradiator (Gamma Services Ltd.) at a dose rate of 0.64 Gy/min. Following irradiation mice received a single tail-vein infusion of 107 BM cells obtained by flushing the femurs of WT and galectin-3^−/− donor mice. Transplanted mice were used 8 weeks posttransplant.

Galectin-3 inhibitor GB1107 (3,4-dichlorophenyl 3-deoxy-3-[4(3,4,5-trifluorophenyl)-1H-1,2,3-triazol-1-yl]-1-thio-α-D-galactopyranoside; Galecto Biotech; ref. 32) was prepared at a concentration of 1 mg/mL in 1% polyethylene glycol, 0.5% hydroxypropyl methyl cellulose (HPMC), and stored in aliquots at −20°C. The anti-PD-L1 monoclonal antibody (clone 10F.9G2) used for in vivo blockade experiments was purchased from BioXCell and 200 μg in PBS was administered twice weekly by interperitoneal injection.

LLC1 cells stably transfected with pCMV-KDEL-Gluc-1 were assessed for luciferase expression upon addition of n-colenterazine (n-CTZ; Lux Biotechnologies 20001) substrate to a final concentration of 10 μmol/L to live cells in 96-well plates. Lymph nodes were disaggregated by passing through 40 μm cell strainers and suspended in PBS. N-CTZ was added at a final concentration of 10 μmol/L. Luciferase activity was assessed with a BioTek SynergyTM HT Luminometer.

Formalin-fixed paraffin embedded sections were deparaffinized in xylene and rehydrated in graded ethanol. Epitopes were retrieved by microwaving in 0.01M sodium citrate (pH 6) for Ym1, and cleaved caspase-3 staining and by proteinase K digest (1.25 mg/mL) for 5 minutes for F4/80 staining. Sections were blocked with serum-free protein block (DAKO) and incubated overnight at 4°C with primary antibodies, rabbit anti-mouse Ym1 (Stem Cell Technologies, 1:200), rat anti-mouse F4/80 (Abcam, 1:100), rabbit anti-mouse ki-67 (Abcam, 1:200), and rabbit anti-mouse cleaved caspase-3 (clone 5A1; Abcam, 1:1,000). Sections were incubated with species-specific biotinylated IgG (Vector), and visualized with 3,3′-diaminobenzidine (DAB) substrate.

Five to ten fields were scored for each tumor representing both tumor and stroma. Absolute cell counts were recorded for F4/80 and Ym1 positive cells. Cleaved caspase-3 staining in tumors was quantified by inverting 8-bit TIFF files so that DAB-positive areas give the highest pixel intensities. Mean pixel intensities (MPI) were then measured in up to 160 fields of view covering the entire tumor parenchyma and averaged to give a single value per tumor.

Tumor sections were incubated with rat anti-mouse F4/80 followed by horseradish peroxidase (HRP)-labeled goat anti-rat IgG (DAKO) and tyramide green (Invitrogen). Sections were microwaved in 0.01M sodium citrate (pH 6) for 5 minutes, reblocked, and probed with rabbit anti galectin-3 (R&D) or rabbit anti Ym1 followed by HRP-labeled goat anti-rabbit IgG (DAKO) and tyramide red (Invitrogen) and mounted in fluoromount-G with DAPI (eBioscience). Images were captured on a Nikon Eclipse E600 microscope.

Total RNA from LLC1 tumors and lung tissue was prepared using RNeasy kits (Qiagen) and reverse transcribed into cDNA using Quantitect RT kits (Qiagen). cDNA was analyzed using either a SYBR green-based quantitative fluorescence method (Invitrogen) and Kiqstart primers (Sigma Aldrich) or Taqman primer probe sets (Life Technologies).

Cells were lysed in NP-40 (Invitrogen) and separated by 10% to 15% SDS-PAGE. Proteins were transferred to nitrocellulose membrane and probed using antibodies against galectin-3, 1:500 (eBioM3/38; eBioscience) and GAPDH, 1:3,000 (14C10; Cell Signaling Technology), followed by species specific HRP-conjugated secondary antibodies (Dako). Bound antibodies were detected using the Enhanced Chemiluminescence 2 Detection Kit (Pierce).

Tumors were minced in serum-free DMEM and digested with Liberase (2 mg/mL; Sigma-Aldrich) and DNase I (Sigma-Aldrich) at 37°C for 30 minutes. Disaggregated tissue was filtered through a 35-μm nylon mesh, washed, and resuspended in FACS buffer (PBS with 0.1% BSA). Fc receptors were blocked with anti-mouse CD16/32 (Biolegend). Antibody cocktails (anti-mouse Ly6G-pacific blue, CD11b-BV605, Galectin-3-FITC, CD45-PerCP and CD45-APCcy7, MHC-II-PE, CD206-PEcy7, PD-1-APC, F4/80-AF700, CD4-pacific blue, PD-L1 BV605, CD3-PerCPcy5.5, IFNγ-PE, CD8-AF700, all from Biolegend) were added to cells and incubated for 20 minutes at room temperature. Samples were fixed and RBCs were simultaneously lysed in RBC Lysis/Fixation solution (Biolegend). For intracellular staining, cells were permeabilized with intracellular staining permeabilization wash buffer (Biolegend) and incubated with anti-CD206 or anti-IFNγ (Biolegend). Cells were analyzed using an LSR-Fortessa cell analyzer (Beckton Dickenson).

Statistical analyses were performed using Graphpad Prism 7.0 software. Results are represented as mean ± SEM and statistical tests are described in the figure legends.

To examine lung cancer growth within the correct tissue compartment, LLC1 cells stably expressing G. princeps-luciferase were injected (1 × 10³ cells) through the intercostal space directly into the lung parenchyma of control and galectin-3^−/− mice. Hematoxylin and eosin staining of lung tissue confirmed the presence of tumors in control but not galectin-3^−/− mouse lungs (Fig. 1A). At 20 days post-injection, 4/10 of control mice had tumors, whereas none of the galectin-3^−/− mice developed tumors (Table 1). In addition, 7/10 control animals displayed gross swelling of the mediastinal lymph nodes (MLN; Table 1), which were positive for metastatic cells as assessed by luciferase assay on homogenized MLNs. Only 1/11 of galectin-3^−/− mice had luciferase positive MLNs (Table 1).

LLC1 cells expressing luciferase were injected subcutaneously in both flanks of WT and galectin-3^−/− mice. After day 10, subcutaneous tumors from control animals were much larger than those of galectin-3^−/− mice. This difference became statistically significant at day 12 (P = 0.0004). By the end of the study, tumors of controls had an average volume of 286 mm³ compared with a volume of 9 mm³ in galectin-3^−/− animals (96.9% reduction, P < 0.0001; Fig. 1B and C). The weight of tumors from control mice was 98% heavier than that of galectin-3^−/− mice, 153 ± 31 mg and 3 ± 2 mg, respectively (P < 0.0001; Fig. 1D). Of all the tumor cell injections received by each group, only 11/40 led to tumors in galectin-3^−/− mice compared with 36/40 in controls (Table 1). A total of 5/12 control mice had luciferase positive metastases in their MLNs whereas galectin-3^−/− mice had no metastases (Table 1). These results indicate that galectin-3^−/− mice do not support tumor establishment and spread in a subcutaneous LLC1 model. Although LLC1 inoculation increased the serum concentrations of anti-galectin-3 IgG antibodies in galectin-3^−/− mice, no correlation, either negative or positive, was established between antibody production and tumor volume (Supplementary Fig. S1).

Tumor stroma F4/80⁺ macrophages were significantly higher in galectin-3^−/− animals compared with control (P = 0.0217). However, the ratio of Ym1⁺/F480⁺ macrophages was significantly higher in controls (P = 0.0484, Fig. 2A and B), indicating that higher galectin-3 levels around the tumor environment can drive expansion of M2 macrophages (28). Transcript analysis from whole tumor RNA showed that control tumors had 2.5-, 3.3-, and 16.7-fold higher levels of IL4, IL10, and IL13 transcripts, respectively (P = 0.04, 0.024, 0.119, respectively), and displayed a 28.8-fold reduction in IFNγ mRNA when compared with galectin-3^−/− tumors (P = 0.0066, Fig. 2C). These results indicate a cytokine environment that favors M2 macrophage activation in tumors of control but not galectin-3^−/− hosts and suggests an important role for galectin-3 in the regulation of TAM phenotype.

We hypothesized that tumor macrophages may contribute to tumor growth in the LLC1 model. C57Bl/6 CD11b-DTR transgenic mice were used as a model of macrophage ablation (33). CD11b-DTR transgenic mice and WT siblings received a single diphtheria toxin (DT) injection immediately prior to cell implant. At day 12, 15/24 tumors developed in CD11b-DTR mice compared with 20/22 in controls. CD11b-DTR animals had significantly smaller tumor volumes compared with controls (29.4 ± 4.1 mm³ and 89.4 ± 0.9 mm³, respectively; P = 0.0005, Fig. 2D) and significantly reduced tumor weights (9.1 ± 1.0 and 23.4 ± 4.0 mg, respectively; P = 0.0011, Fig. 2E). To assess the efficiency of macrophage ablation in this model, DT was administered to mice with established tumors and F4/80 staining carried out 24 hours after DT administration. An 88% reduction in TAMs was observed in the tumors of DTR transgenic animals (P < 0.0001; Supplementary Fig. S2A).

To determine which galectin-3–expressing cells are necessary to support tumor growth, we first altered galectin-3 expression in recruited cells. Control and galectin-3^−/− mice were irradiated and transplanted with 10⁷ control or galectin-3^−/− BM cells. Eight weeks post-BM transplant, LLC1 cells were injected subcutaneously. Transplantation of control BM cells into galectin-3^−/− mice resulted in significantly increased average tumor volume and final tumor weight compared with mice transplanted with galectin-3^−/− BM cells (final tumor volume 336 mm3 and final weight of 297.6 mg compared with 163.9 and 124.4 mg, P < 0.0001 and P = 0.0007, respectively; Fig. 2F and G). Dual immunofluorescence staining showed that the stroma of tumors harvested from galectin-3^−/− animals transplanted with control BM had F4/80 and galectin-3 dual positive cells (Fig. 2H), although the total number of infiltrating macrophages was not different between control or galectin-3^−/− BM transplanted mice (Supplementary Fig. S2B), suggesting that galectin-3 positive macrophages are recruited to the tumor stroma and contribute to tumor growth.

LLC1 cells in vitro display cell surface and cytoplasmic galectin-3 staining and release galectin-3 into the culture medium (Supplementary Fig. S3A). To determine whether tumor-derived galectin-3 also contributes to tumor growth, galectin-3 was stably knocked down (KD) in LLC1 cells prior to subcutaneous injection (Supplementary Fig. S3B). Although LLC1 proliferation was reduced by galectn-3 KD in vitro (Supplementary Fig. S3C and S3D), tumor growth and final tumor weights of LLC1-galectin-3-KD cells was similar to WT cells (Supplementary Fig. S3E–S3G).

Recently, a series of monosaccharide galectin-3 inhibitors with low nmol/L affinities and good selectivity over other galectins have been described (32). From this series GB1107 has high affinity in man at 37 nmol/L but due to species differences in the galectin-3 carbohydrate binding domain (CBD), the mouse galectin-3 affinity is 38-fold lower. GB1107 has low clearance (1.2 ml/min/kg, t_1/2 = 4.5 hours, i.v.) and good uptake upon oral administration, resulting in high oral availability (F = 75%, orally). As a consequence, dosing GB1107 at 10 mg/kg orally once daily results in a plasma concentration above mouse K_d over 24 hours (Supplementary Fig. S4). CD-1 nude mice bearing human lung A549 adenocarcinoma xenografts were treated from day 18 postimplantation once daily with 10 mg/kg GB1107. This resulted in significantly reduced tumor growth and final tumor weights (46.2% smaller compared with vehicle control tumors with final average weights of 117 ± 16 and 63 ± 11 mg, respectively; P = 0.0132; Fig. 3A). Treatment with GB1107 also inhibited LLC1 growth (tumor volumes decreased 48% compared with controls on day 18, P < 0.001) and reduced final tumor weights (47 ± 14 mg vs. 120 ± 29 mg controls, P = 0.0524) when administered daily from the outset (Fig. 3B). Transcript analysis of tumor RNA from the LLC1 tumors revealed reduced galectin-3 (48% less than vehicle, P = 0.018) and mesenchymal markers TGFβ (45% less than vehicle, P = 0.015) and trends for reductions of VEGF and αSMA expression (Fig. 3C). Although there was no change in the M1 marker Nos2, there was a trend towards a reduction in expression of the M2 marker Ym1 (50% less than vehicle) and CD98 (49% less than vehicle), which drives galectin-3–mediated M2 macrophage activation (28), suggesting a decrease in M2 skewed TAM accumulation in the tumor. To examine effects on metastasis, LLC1 cells were injected intravenously and lung colonization was determined at 7 days postinjection (Fig. 3D). The presence of metastasis was examined by expression of Gaussia luciferase transcript in whole lung RNA extracts. GB1107 administered daily from day 1 significantly reduced tumor burden by 79.2%. These data suggest that inhibition of galectin-3 with an orally active selective galectin-3 inhibitor can significantly reduce lung adenocarcinoma growth and metastasis in vivo.

Mice were generated that express the human LGALS3 gene in place of the mouse gene (Hu-Gal-3-KI). Western blot analysis confirmed expression of only human galectin-3 in mouse liver lysates from Hu-Gal-3-KI mice (Fig. 3E). LLC1 tumor growth was inhibited by GB1107 in Hu-Gal-3-KI mice when administration was delayed until day 5 after inoculation (Fig. 3F).

Given the altered M1:M2 TAM ratio in LLC1 tumors from galectin-3^−/− mice and inhibitor-treated mice, we next determined the role of LLC1-derived galectin-3 on macrophage polarization. Conditioned media from LLC1 cells in vitro increased IL4-stimulated arginase activity in BM-derived macrophages (BMDM) and increased gene expression of arginase-1 and fizz1 (Supplementary Fig. S5A and S5B). This increase was inhibited by GB1107, suggesting galectin-3 secreted by LLC1 cells induces macrophages to adopt an alternative M2-like phenotype (Supplementary Fig. S5A and S5B). GB1107 did not affect LPS-induced Nos2 expression or nitric oxide (NO) production by BMDMs (Supplementary Fig. S5A and S5C). Although our data show TAMs to be a vital determinant of tumor growth in vivo, treatment of LLC1 cells with inhibitor in vitro also impacted on cell migration and proliferation albeit at higher concentrations (Supplementary Fig. S5D–S5F), suggesting some direct effect on galectin-3-mediated tumor cell expansion and migration.

TAMs can contribute to T-cell immunosuppression (4). In particular, M2-like macrophages secrete more galectin-3 (34) and galectin-3 directly impedes T-cell infiltration and activation (24–27, 35). We therefore investigated whether galectin-3–dependent M2 polarization is also associated with changes in T-cell infiltration and activation in vivo. Flow cytometric analysis of tumor digests from hu-Gal-3-KI mice treated with GB1107 showed no increase in macrophage infiltration, but showed a decrease in macrophage CD206 expression indicative of reduced M2 TAMs (Supplementary Fig. S6A; Fig. 4A). Similarly, although there was no significant change in the total number of CD3⁺ T cells, GB1107 caused an increase in CD8⁺ but not CD4⁺ T cells within tumors (Fig. 4A). This pattern of immune infiltration was also observed in tumor digests from galectin-3^−/− mice compared with WT C57Bl/6 mice (Fig. 4B). Moreover, in galectin-3^−/− tumors, infiltrating CD8 T cells but not CD4⁺ T cells displayed trends toward increased PD-1 and IFNγ expression, together suggesting that galectin-3 depletion may reprogram the tumor microenvironment to favor proinflammatory M1-like macrophages and enhance cytotoxic CD8 T-cell infiltration and activation. Galectin-3^−/− mouse tumors displayed no overall changes in total CD45⁺ cells, neutrophils, monocytes, or dendritic cells (DC) compared with WT tumors (Supplementary Fig. S6B).

To assess whether other systemic changes in galectin-3–depleted mice may also influence CD8 T-cell activation, we assessed myeloid populations within the BM of WT and galectin-3^−/− mice. We observed no differences in total CD11b⁺ cells, neutrophils, or monocytes including inflammatory Ly6C^hi monocytes (Supplementary Fig. S7). There were however increased DCs in galectin-3^−/− compared with WT mice, suggestive of another indirect mechanism by which galectin-3 may enhance antitumor T-cell priming (26).

We next examined the effect of galectin-3 inhibition in combination with immune checkpoint inhibition. In this study, GB1107 treatment was delayed until day 6 postimplantation. Delayed administration of GB1107 alone did not reduce tumor burden and administration of an anti-PD-L1 antibody administered twice weekly intraperitoneally from day 6 also had no impact on tumor growth. However, a combination of GB1107 and anti-PD-L1 antibody treatment significantly potentiated the effect of the single agents (49.5% and 51.4% reduced tumor volumes and weights respectively compared with untreated controls; Fig. 5A and B). The reduced tumor growth in the combination group was associated with an increase in PD-1⁺ CD8⁺ T cells (Fig. 5C; Supplementary Fig. S8A–S8B) and reduced CD206 expression in macrophages (Fig. 5C). This was combined with a significant decrease in tumoral αSMA (Supplementary Fig. S8C) and an increase in expression of T-cell cytotoxic mediators (IFNγ perforin-1, granzyme B, and fas ligand; Fig. 5D) and a 10.3% increase (P = 0.005) in the apoptosis marker cleaved caspase-3 (Fig. 5E and F) but was not associated with changes in tumor cell proliferation as determined by Ki-67 IHC staining (Supplementary Fig. S8D and S8E). There was a corresponding reduction in galectin-3 protein levels within GB1107-treated tumors (Supplementary Fig. S8F). Together our data suggest that combination therapy with galectin-3 inhibitor GB1107 and PD-L1 blocking antibody promotes tumor cell apoptosis and cytotoxic CD8 T-cell activation.

In this study we show that galectin-3^−/− mice do not support LLC1 tumor growth and deleting galectin-3 in BM-derived cells recruited to engrafted mouse lung adenocarcinomas inhibits tumor growth and spread despite high expression of galectin-3 in the tumor cells. Furthermore, macrophage depletion reduces monocyte recruitment and LLC1 tumor growth, confirming that depletion of macrophages with liposomal clodronate inhibits LLC1 tumor growth (36). Although the spleen can be an important source of TAMs in a KRAS and P53-driven model of lung adenocarcinoma (37), the BM compartment has been shown to be the major source of TAMs in the LLC1 model (38). Therefore, we sought to restore galectin-3 in tumor macrophages in galectin-3^−/− mice by BM transplant with WT galectin-3-positive BM cells. This results in an increase in tumor growth similar to that observed in WT mice. Our previous work has shown that galectin-3 is an important regulator of macrophage function, promoting an “M2” phenotype (28). Our data show that macrophages in tumors from galectin-3^−/− mice or mice treated with GB1107 have reduced CD206⁺ M2-like macrophages and we observe reduced M2-promoting cytokine transcripts and elevated IFNγ expression within galectin-3^−/− tumors. In addition, conditioned media from LLC1 cells increases alternative activation of macrophages in vitro and this can be blocked by coculture with GB1107. This demonstrates that galectin-3 contributes to the M2 immunosuppressive function of TAMs.

TAMs promote many important features of tumor progression including angiogenesis, tumor cell invasion, motility, and metastasis and can also suppress T-cell responses (4). These data show that galectin-3–expressing macrophages are recruited to the tumor site, develop an M2 phenotype and induce downregulation of CD8⁺ CTL functions. Galectin-3 has been shown to induce T-cell tolerance resulting in T-cell anergy, through various mechanisms including inhibiting CD8 and TCR clustering (39), destabilizing the immune synapse and promoting internalization of TCR and CD3ζ chains (40). It can also restrict membrane movement and TCR-associated signaling functions of CD45 (41) and inhibit LFA-1 recruitment thus disrupting proper secretory synapse formation and secretion of IFNγ (27).

Galectin-3 may also suppress CTL effector function by binding to LAG-3, a negative regulatory checkpoint, on CD8⁺ T cells (26) and by inducing apoptosis of CTLs (25) and impairs the antitumor functions of natural killer (NK) cells (42). CTLs activated in vitro show an alteration in the N-glycome with longer and more branched N-glycans resulting in the expression of surface glycoproteins that exhibit high galectin-3 binding (43). The high concentration of galectin-3 found in tumor microenvironments could potentially explain the loss of CTL functions through reduced motility and signaling functions of surface molecules.

Galectin-3^−/− mice have also been shown to have an increase in lymph node plasmacytoid DCs (pDC) compared with WT mice, which are superior in activating CD8+ CTLs compared with conventional DC (26). In addition, galectin-3 knockdown in monocyte-derived DCs increases the proliferation and IFNγ production from antigen-stimulated CD4⁺ T cells (44). Our profiling of BM from WT and gal-3 KO mice showed an increase in CD45⁺/MHC-II⁺/CD11b- DCs in BM of gal-3 KO mice compared with WT. Although our study did not distinguish DC subsets, together the data suggest that galectin-3 may indirectly regulate CD8 function by promoting DC functions. This requires further study.

We show that treatment with GB1107 alone from the outset inhibits LLC1 growth and delayed treatment inhibits LLC1 growth in human galectin-3–expressing mice. This reflects the increased affinity this inhibitor has on human versus mouse galectin-3. In addition, the galectin-3 inhibitor significantly potentiates the effect of immune checkpoint blockade with an anti-PD-L1 blocking antibody. It is believed that the limited patient responses to checkpoint inhibition is attributable to the lack of T-cell infiltration in so-called “cold” tumors (2). Gordon-Alonso and colleagues, show that galectin-3 binds to the extracellular matrix and to glycosylated IFNγ, preventing release of IFNγ-induced CXCL9, which acts as a T-cell chemo-attractant (35). Consistent with this, GB1107 both alone or in combination with anti-PD-L1 increases the number of tumor infiltrating CD8 CTLs. Therefore, galectin-3 inhibition might provide the critical means to turn a “cold” tumor “hot,” and thus responsive to immune checkpoint intervention. Furthermore, CD8⁺ CTLs within the combination drug-treated tumors are more activated (express more surface PD-1), and the cytokine environment favors tumor rejection with increased expression of cytotoxic (IFNγ, perforin-1, and granzyme B) and apoptotic (fas ligand) genes with increased caspase activation.

Reduced galectin-3 expression within tumor cells has been shown to reduce tumor growth in many cancers (reviewed in ref. 45), suggested to be due to the anti-apoptotic effect of cytoplasmic galectin-3 binding to K-RAS and engaging anti-apoptotic pathways via its NWGR motif (20). However, we show that knockdown of galectin-3 in tumor cells with shRNA had only a partial effect on tumor growth in vitro but had no significant effect on LLC1 growth in vivo. We also show that treatment with the galectin-3 inhibitor alone could inhibit human adenocarcinoma growth in CD-1 nude mice, which lack a T-cell response but which display innate immunity. These suggest that either tumor derived or macrophage-derived galectin-3 can impact on tumor growth in this model, independent of the T-cell–mediated effects.

In conclusion, our results demonstrate that galectin-3 inhibition leads to a reduction in M2-like TAMs and increased infiltration and activity of CD8⁺ CTLs within LLC1 tumors resulting in reduced tumor growth and metastasis. Several studies have used other approaches to inhibit galectin-3 in cancer including peptide inhibitors (46), lactulose amines (47), a glycopeptide isolated from cod (48) and large complex plant-derived polysaccharides including modified citrus pectin (49), GCS-100 (39), and galactomannans such as GM-CT-01 (50). GCS-100 is currently being developed for chronic lymphoid leukemia and multiple myeloma (51). However recent evidence suggests that these complex carbohydrates do not act as inhibitors of the canonical carbohydrate-binding site of galectin-3 and their physiologic effects may be due to unrelated actions (52). We show using a specific and high affinity inhibitor of the galectin-3 carbohydrate site that pharmacologic inhibition of galectin-3 inhibits lung adenocarcinoma growth and potentiates the effect of immune checkpoint inhibitors. Therefore, galectin-3 has a strong regulatory effect on cancer-related inflammation and could present a key target in the management of lung, and potentially other galectin-3–driven carcinomas, in combination with immune checkpoint blockade.

L. Vuong reports receiving other commercial research support from Galecto Biotech. N.C. Henderson is a consultant/advisory board member of Galecto Biotech. U.J. Nilsson has ownership interest (including stock, patents, etc.) in Galecto Biotech AB and is a consultant/advisory board member of Galecto Biotech AB. H. Leffler reports receiving a commercial research grant from Galecto Biotech AB, has ownership interest (including stock, patents, etc.) in Galecto Biotech AB, and is a consultant/advisory board member of Galecto Biotech AB. H. Schambye is a CEO at Galecto Biotech and has ownership interest (including stock, patents, etc.) in Galecto Biotech. T. Sethi is a Chief Physician Scientist at AstraZeneca and has ownership interest (including stock, patents, etc.) in Galecto Biotech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Vuong, E. Kouverianou, S.E.M. Howie, F.R. Zetterberg, U.J. Nilsson, H. Leffler, H, Schambye, T. Sethi, A.C. MacKinnon

Development of methodology: L. Vuong, E. Kouverianou, U.J. Nilsson, S. Tantawi, A.C. MacKinnon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Vuong, E. Kouverianou, C.M. Rooney, B.J. McHugh, C.D. Gregory, A. Pedersen, L. Gravelle, A.C. MacKinnon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Vuong, E. Kouverianou, B.J. McHugh, S.E.M. Howie, S.J. Forbes, F.R. Zetterberg, U.J. Nilsson, H. Leffler, P. Ford, H, Schambye, A.C. MacKinnon

Writing, review, and/or revision of the manuscript: L. Vuong, E. Kouverianou, B.J. McHugh, S.E.M. Howie, C.D. Gregory, N.C. Henderson, F.R. Zetterberg, U.J. Nilsson, H. Leffler, P. Ford, A. Pedersen, H, Schambye, T. Sethi, A.C. MacKinnon

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.J. Forbes

Study supervision: T. Sethi

Other (design and synthesis of GB1107): F.R. Zetterberg

We thank Frank McCaughan for helpful discussions and kind provision of reagents, as well as the Biomedical Research Council Flow Cytometry Core (King's College) and the University of Edinburgh Queen's Medical Research Institute's Flow Cytometry facility for flow sorting and flow cytometric analysis. The work presented was funded by Galecto Biotech, a King's Health School's scholarship to L. Vuong, and a Norman Salvesen Emphysema Research grant to E. Kouverianou.

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.