Targeting galectin-driven regulatory circuits in cancer and fibrosis

Galectins are a family of endogenous glycan-binding proteins that have crucial roles in a broad range of physiological and pathological processes. As a group, these proteins use both extracellular and intracellular mechanisms as well as glycan-dependent and independent pathways to reprogramme the fate and function of numerous cell types. Given their multifunctional roles in both tissue fibrosis and cancer, galectins have been identified as potential therapeutic targets for these disorders. Here, we focus on the therapeutic relevance of galectins, particularly galectin 1 (GAL1), GAL3 and GAL9 to tumour progression and fibrotic diseases. We consider an array of galectin-targeted strategies, including small-molecule carbohydrate inhibitors, natural polysaccharides and their derivatives, peptides, peptidomimetics and biological agents (notably, neutralizing monoclonal antibodies and truncated galectins) and discuss their mechanisms of action, selectivity and therapeutic potential in preclinical models of fibrosis and cancer. We also review the results of clinical trials that aim to evaluate the efficacy of galectin inhibitors in patients with idiopathic pulmonary fibrosis, nonalcoholic steatohepatitis and cancer. The rapid pace of glycobiology research, combined with the acute need for drugs to alleviate fibrotic inflammation and overcome resistance to anticancer therapies, will accelerate the translation of anti-galectin therapeutics into clinical practice. Galectins are glycan-binding proteins with broad roles in physiological and pathological processes. Here, Rabinovich and colleagues discuss the therapeutic relevance of galectins to cancer and fibrosis and review the various strategies to target galectins and their progress in clinical trials.


Introduction
The complex repertoire of glycan structures present in cells and tissues (that is, the glycome) stores crucial biological information that contributes to the reprogramming of cellular fate and function and thus has a profound influence on the delicate balance between health and disease [1][2][3][4] . The diversity and spatiotemporal regulation of glycans within glycoconjugates rely on the synchronized action of glycan-modifying enzymes, including glycosyltransferases and glycosylhydrolases. Glycan remodelling is regulated by intracellular and environmental signals, such as metabolic stress, oxygen and nutrient availability, growth factors and cytokines 2,5 . Dysregulation of many cellular processes, including cellular communication, proliferation, differentiation and survival has been linked to aberrant glycosylation. This finding implicates the glycome in the pathophysiology of nearly every major disease, with notable impact on cancer, inflammation and fibrosis 5,6 . Changes in the glycosylation signature of tumour, immune and endothelial cells (ECs) are among the common hallmarks of the tumorigenic process, as these signatures can influence cell adhesion, epithelial-to-mesenchymal transition (EMT), angiogenesis, immunoediting and metastasis 1,5,[7][8][9][10][11][12] . Moreover, selective glycan profiles may also help to control the initiation, persistence and resolution of inflammatory and fibrotic processes 1,13 . Thus, an aberrant glycome might alter cellular functions by regulating the exposure or masking of specific glycoepitopes. These aberrations can ultimately lead to the development of pathological responses. Converting glycan-encoded information into biological programmes relies, at least in part, on the contributions of endogenous glycan-binding proteins or lectins 6 . The three major lectin families that have decisive roles in shaping both inflammatory and tumour microenvironments are sialic acid-binding immunoglobulin-like lectins (Siglecs), C-type lectin receptors (including selectins 14 ) and galectins. Galectins are a family of soluble lectins that share affinity for β-galactoside-containing saccharides 15 . Examples of some natural ligands for mammalian galectins are summarized in Box 1. Galectins might be involved in the transition from healthy to neoplastic or inflamed tissues and contribute to the persistence of these pathological conditions via both intracellular and extracellular mechanisms 16,17 . Galectins also influence most hallmarks of tumour progression 17 and modulate resistance to numerous anticancer treatments, including immunotherapy, chemotherapy, radiotherapy, targeted therapies and anti-angiogenic therapy 18 . Furthermore, galectins might contribute to regulatory circuits that amplify, sustain or alleviate tissue fibrosis and inflammation by selectively targeting different cell types and their microenvironments 16,[19][20][21] . Accordingly, these glycan-binding proteins have been proposed as therapeutic targets in a broad range of pathological conditions and are currently under clinical evaluation.
In this Review, we consider the pathophysiological relevance of specific galectin-glycan interactions from a translational perspective and discuss the design, mechanisms of action, selectivity and therapeutic relevance of galectin-targeted agents. We focus on the results of preclinical studies and clinical trials and highlight lessons learned from targeting galectins in a diverse group of disease states, including nonalcoholic steatohepatitis (NASH) 19 , idiopathic pulmonary fibrosis (IPF) [22][23][24] , as well as various malignancies 11,17 . These examples underscore the numerous potential opportunities to capitalize on galectin-glycan interactions for therapeutic purposes.
We focus in particular on agents designed to target galectinglycan interactions, including small-molecule inhibitors, natural polysaccharides and their synthetic derivatives, as well as peptides

Review article
The responses generated by galectin-glycan interactions -typically resulting from intracellular and extracellular activities -modulate a broad range of cellular responses, including proliferation, differentiation and survival 60 .
Dysregulated galectin expression occurs in a range of malignancies and is frequently associated with poor prognosis and limited responses to anticancer therapies 17,18,61,62 . Moreover, galectin levels are frequently elevated at various stages of the fibrotic cascade 21,63 . These
The altered glycosylation profile observed in tumour cells led to the definition of tumour-associated carbohydrate antigens (TACAs). Depending on the cell type, N-glycans can be altered by increasing synthesis of poly-LacNAc chains or β(1,6) branching, which are permissive modifications for galectin binding. In turn, mucin-type O-glycans can be truncated and consequently, tumour cells express higher levels of structures such as the Thomsen-Friedenreich antigen (Tf antigen), which may serve as galectin ligands 5,11,61 .

GAL3 GAL9
Review article data highlight the emerging roles of galectins as possible biomarkers and therapeutic targets in these pathological conditions. In the next sections we discuss the role of galectins in cancer and fibrosis, with a focus on GAL1, GAL3 and GAL9.

Galectins in cancer
Galectins have a broad influence on tumour progression via glycosylation-dependent or independent mechanisms that modulate proliferation, evasion of growth suppressors and immune responses, resistance to cell death, induction of angiogenesis, inflammation and metastasis 17,64 . These galectin-influenced circuits can shape tumour, endothelial and immune compartments in individual tumour microenvironments (Fig. 2). These actions have both local and systemic impact and might contribute to processes that influence clinical outcomes in patients with cancer. To substantiate the therapeutic relevance of these glycan-binding proteins, we provide several examples that highlight the impact of galectins, most notably GAL1, GAL3 and GAL9, on the reprogramming of the tumour, vascular and immune landscapes and their contributions to tumour growth and progression (Fig. 2).

Galectins reprogramme tumour cell fate
GAL1 is upregulated in most tumour tissues and their associated stroma and has been proposed as a biomarker of poor prognosis in breast, colon, lung, prostate adenocarcinoma and melanoma 65 . Tumours of the digestive tract and urinary system, as well as thyroid cancer and melanomas, express high levels of GAL3; by contrast, GAL3 expression is downregulated in tumours of the reproductive tract. Subcellular compartmentalization of GAL3 is a crucial factor contributing to its pro-or antitumoural roles. For example, increased GAL3 within the nuclear compartment has been associated with a better outcome among patients with neuroblastoma 66 . On the other hand, GAL9 expression can be up-or downregulated in association with neoplastic transformation depending on the specific tumour type 65,67 . As the expression of these lectins can change during tumour progression, they have been proposed as biomarkers for diagnosis, prognosis and/or intervention in various cancer stages 65 . Extracellular GAL1 and GAL3 directly modulate tumour cell fitness, migration, EMT and stemness through their interactions with glycosylated tumour-associated receptors, including epidermal growth factor (EGF), transforming growth factor β (TGFβ) receptors and cell surface integrins [68][69][70][71][72][73][74] , that contribute to cancer progression and metastasis. GAL1 can trigger EMT in gastric cancer 75 and hepatocellular carcinoma 72,76 via non-canonical activation of the Hedgehog pathway 75 , downregulation of E-cadherins 72 and induction of α v β 3 integrin-dependent AKT signalling 76 . Collectively, these actions have an impact on clinical outcomes and therapeutic responses to both sorafenib and doxorubicin 77,78 . GAL1 also regulates events that promote the growth of pancreatic adenocarcinoma, including tumour cell proliferation, invasion and metastasis 78 , and GAL1 silencing inhibits migration and invasion of metastatic castration-resistant prostate cancer cells via suppression of androgen receptor and AKT-mediated signalling 79 . In turn, interactions between GAL3 and the glycosylated transmembrane protein mucin 1 (MUC1) enhance EGF receptor dimerization and activation (Fig. 2), and thus promote proliferation and motility of epithelial cancer cells by activating ERK1/2 and AKT signalling pathways 80,81 . Notably, complex N-glycans linked to EGF receptors on breast carcinoma cells control EMT, cell motility and tumour metastasis by facilitating the exposure of GAL3-specific glycoepitopes 68 GAL1, 2,5,7,10,11,13,14,15 Prototype a GAL3 Chimera-type b GAL4, 6,8,9,12 Tandem repeat-type c Glycosylated receptors

Fig. 1 | Structural classification of galectins.
Galectins are classified into three different subfamilies on the basis of their structure: a, prototype galectins (GAL1, GAL2, GAL5, GAL7, GAL10, GAL11, GAL13, GAL14 and GAL15) that function as monomers or non-covalent homodimers with identical carbohydrate recognition domains (CRDs); b, chimera-type galectins (GAL3) that contain only one CRD and have the capacity to oligomerize via a non-lectin N-terminal domain; and c, tandem repeat-type galectins (GAL4, GAL6, GAL8, GAL9 and GAL12) that contain two CRDs each with a distinct specificity joined by a flexible linker peptide. Depending on distinct carbohydrate-binding specificities and oligomerization status, these proteins can form lattices by interacting with multiple glycosylated ligands on the cell surface and/or with extracellular matrix glycoproteins. In doing so, they elicit different biological responses via mechanisms that include assembly of multivalent glycan-galectin complexes and modulation of transmembrane signalling events by selective receptor segregation, retention and endocytosis.
Review article (Fig. 2). Tumour-derived GAL3 also promotes invasion via its capacity to impair interactions between adhesion molecules expressed on the surface of malignant cells and N-glycosylated proteins within the extracellular matrix, such as laminin and fibronectin 74 . Furthermore, GAL3 promotes the establishment of metastatic niches by binding to the Thomsen-Friedenreich antigen (Tf antigen) expressed by metastatic lung tumour cells 70 and facilitating homotypic and heterotypic aggregation and emboli formation 82 . Similarly, overexpression of GAL3 promotes proliferation, migration and invasion of oral squamous cell carcinoma (OSCC) cells via enhanced WNT-β-catenin signalling and EMT 83 . The interaction of GAL3 with α v β 3 integrin promotes KRAS addiction 84 ; these findings identified GAL3 as a potential druggable target for 'KRAS-addicted' lung and pancreatic cancers (Fig. 2).
GAL1 and GAL3 also promote oncogenic signalling, apoptosis resistance and cell cycle progression when present in the intracellular compartment 85,86 . GAL1 associates with oncogenic HRAS (Fig. 2) and enhances HRAS membrane anchorage and oncogenesis by increasing signalling to the kinase RAF1 (refs. 87,88 ). Likewise, GAL3 promotes cancer cell proliferation and survival via its interactions with activated KRAS and enhances colorectal tumour cell invasion via constitutive activation of RAF-MEK-ERK signalling 88,89 (Fig. 2). Interestingly, in vitro studies revealed that a combination of the KRAS inhibitor salirasib and the GAL3 inhibitor modified citrus pectin (MCP) induced cell cycle arrest and apoptosis in anaplastic thyroid cells 90 . Other studies showed that GAL3 expression was suppressed by p53. This response was largely eliminated in cells with TP53 mutations, thus explaining the elevated levels of GAL3 expression in TP53-mutant tumours 91 . Remarkably, galectin-mediated mechanisms also promote tumour resistance to most anticancer therapies 18 , including chemotherapy 77 , immunotherapy 92 , anti-angiogenic therapy 9 , radiotherapy 93 and targeted therapies 94 .

Galectins induce aberrant angiogenesis
Abnormal angiogenesis and diversion of pre-existing blood vessels are crucial in neoplastic as well as inflammatory and fibrotic conditions, as these processes provide the nutrients and oxygen needed to sustain otherwise hypoxic environments 95 . Interestingly, the glycome of ECs varies considerably in response to immunosuppressive or hypoxic stimuli, which favour exposure of galectin-specific glycan epitopes 9 . Several galectin family members have pro-angiogenic activity as a result of their capacity to bind distinct glycosylated receptors on the EC surface 96 . For example, GAL1 interacts with complex N-glycans on vascular endothelial growth factor (VEGF) receptor 2 (VEGFR2) and promotes receptor phosphorylation followed by activation of the AKT and ERK1/2 signalling pathways, thus preserving vascularization in tumours that are resistant to anti-VEGF treatment 9 (Fig. 2). These actions link tumour hypoxia to aberrant angiogenesis pathways identified in melanoma as well as lung, prostate and pancreatic adenocarcinoma 9,78,97-101 . Likewise, interactions with the transmembrane glycoprotein, neuropilin 1 (NRP1) are essential for GAL1-driven angiogenesis and induction of vascular permeability 102,103 (Fig. 2). Recently, an alternative mechanism was proposed whereby GAL1 regulates angiogenesis through binding to the mRNAs of angiogenesis-related factors, including VEGF-A, early growth response 1 protein (EGR1) and α 5 laminin 104 .
GAL3 promotes vascularization via interactions with complex branched N-glycans present on α v β 3 integrin and VEGFR2 (Fig. 2) that promote retention of these receptors on the EC surface 105,106 . Interestingly, this pro-angiogenic effect involves cleavage of the N terminus of GAL3 by matrix metalloproteinase 2 (MMP2) and MMP9 (ref. 107 ). By contrast, GAL9 mediates angiogenesis through different mechanisms 108 . In vitro studies with individual GAL9 CRDs revealed opposing anti-and pro-angiogenic effects, suggesting a dual role for this lectin in regulating vascular networks 109 . Finally, in addition to their role in promoting angiogenesis, interactions of GAL3 with ECs stimulate the secretion of metastasis-promoting cytokines 110 and the expression of cell adhesion molecules 111 . Thus, agents that target galectins and/or their glycosylated receptors may prevent aberrant vascularization, interrupt the metastasis cascade and circumvent resistance to anti-angiogenic therapies.
The role of galectins in suppressing antitumour immunity involves multiple pathways linked to innate and adaptive immune

Multivalency and ligand density in galectin recognition
The weak binding interaction between lectins and carbohydrate moieties can be compensated by a multivalent display of residues at the cell surface. This is known as the 'cluster glycoside effect', which refers to the fact that multivalent (as opposed to monovalent) interactions will result in an overall increase in binding affinity on a per mole or valence-corrected basis. This is one of the key factors regulating the biological activities of galectins at the cell surface. For example, galectin-permissive glycoepitopes can be exposed differentially, largely depending on the nature of the parent glycoconjugate, the target cell type, the cellular activation status and the underlying physiological or pathological conditions. Likewise, extracellular galectins can form multimers that will cross-link specific cell surface glycoconjugates, thereby generating galectin-glycan complexes or lattices that can regulate one or more intracellular signalling pathways. However, not every glycoconjugate with a permissive glycosylation profile can function in this role. A functional glycoconjugate must be presented in the proper orientation and with appropriate density. Therefore, receptor expression and glycosylation, ligand density and glycan geometry are crucial factors that control multivalent galectin binding, signalling and ultimately biological responses. A more thorough understanding of galectin-glycan lattices will be needed to support the design of multivalent glycosylated architectures, which is currently a highly active field of research in medicinal chemistry. High-affinity ligands for lectins, including those that mimic multimeric interactions identified in natural systems have been designed, synthesized and tested. Synthetic multivalent ligands typically include a central scaffold or core structure, a spacer or functionalized linker and lectin-binding glycoepitopes. Scaffolds with diverse properties, structure, flexibility and valency have been used to prepare multivalent ligands for biomedical applications.
To add further complexity, galectins can also impair T cell activation by forming multivalent lattices that restrict the mobility of relevant immune receptors 50,138 . For example, GAL3 dampens antitumour They interact with receptors such as mucin 1 (MUC1), epidermal growth factor receptor (EGFR) and α v β 3 integrin and with intracellular proteins such as RAS to modulate signalling pathways that promote epithelialmesenchymal transition (EMT), migration and invasion. b, Galectins control endothelial cell responses and promote aberrant angiogenesis by engaging various glycosylated cell surface receptors including neuropilin 1 (NRP1; ligand for GAL1), vascular endothelial growth factor receptor 2 (VEGFR2; ligand for GAL1 and GAL3) and α v β 3 integrin (ligand for GAL3), thereby modulating cell surface retention, endocytosis and signalling. GAL1 also modulates angiogenesis by binding to the mRNA of angiogenesis-related factors such as early growth response protein 1 (EGR1). c, Galectins shape the immune landscape of the tumour microenvironment by acting on both lymphoid and myeloid cells. Additionally, they suppress antitumour responses by engaging immune checkpoint or co-inhibitory molecules including CD45 and CD43 (GAL1), CTLA4 and LAG3 (GAL3) and PD1 and TIM3 (GAL9) leading to inhibition of lymphocyte cell-specific protein-tyrosine kinase (LCK)-mediated T cell activation, or promotion of STAT3-mediated myeloid cell reprogramming. Therefore, galectins facilitate tumour immune escape and mediate resistance to immunotherapeutic strategies. BAT3, HLA-B-associated transcript 3; ECAD, E-cadherin; GLI1, glioma-associated oncogene homologue 1; MMP, matrix metalloproteinase; PP2A, protein phosphatase 2 phosphatase activator; PTP, protein tyrosine phosphatase; SHP2, protein tyrosine phosphatase non-receptor type 11.

Review article
responses by decreasing T cell receptor (TCR) mobility through formation of a glycoprotein lattice that limits interactions between the TCR and CD8 coreceptor, thus promoting dysfunction and anergy of tumour-infiltrating lymphocytes 138 (Fig. 2). Interestingly, galectins may also support transmission of immune inhibitory signals by acting as ligands of co-inhibitory and immune checkpoint molecules, illustrated by the association of GAL1 with CD45 and CD43 (ref. 131 ), GAL3 with LAG3 and CTLA4 (refs. 139,140 ) and GAL9 with TIM3 and PD1 (refs. 133,141 ) (Fig. 2). By interrupting immune-activating mechanisms or triggering immune inhibitory pathways, galectins thwart various immunotherapeutic modalities. For example, GAL1 can reprogramme the tumour endothelium to upregulate both PDL1 and GAL9; blockade of this pathway increased T cell infiltration into the tumour and improved the response to anti-PD1 therapy 92 . Moreover, GAL9-mediated cross-talk between PD1 and TIM3 controls T cell exhaustion programmes and sensitivity to anti-PD1 therapy 133 . Thus, targeting galectins and/or their glycosylated ligands might counteract both local and systemic immunosuppressive circuits, dismantle tumour immune escape mechanisms and recalibrate responses to immunotherapeutic modalities.

Galectins in fibrosis
Pathological fibrosis is the consequence of abnormal mechanisms of tissue repair that are typically associated with cellular stress, chronic inflammation and/or severe tissue damage. It can lead to organ failure. Inadequate tissue regeneration has been described in several chronic fibroproliferative diseases such as IPF 24 and chronic inflammatory diseases such as NASH 19 . The abnormal wound-healing process results in scar formation with excessive collagen deposition via mechanisms that involve fibroblasts, epithelial and endothelial cells, and immune cells, notably macrophages 19 . Given the high prevalence of fibrotic diseases and the limitations of current drugs, which mainly limit the rate of organ dysfunction, the quest for effective therapies is clearly important.
Similarly to GAL3, GAL1 and GAL9 are also thought to contribute to the fibrotic process, largely on the basis of studies in galectin-deficient mice 152,153 . Increased levels of GAL1 mRNA were detected in the lungs of patients with IPF 154 , and GAL1 and GAL9 were prominently expressed in fibrotic liver 155,156 . The pro-fibrotic role of GAL1 is associated with TGFβ-driven fibroblast differentiation 157 . Accordingly, silencing GAL1 prevented hypoxia-induced pulmonary fibrosis and blocked the anticipated decline in lung function 154 . The role of GAL9 in fibrotic diseases of the lung remains under debate. Matsumoto et al. 158 proposed that GAL9 may have a protective role on the basis of its ability to suppress growth and induce dose-dependent apoptosis of human lung fibroblasts. However, GAL9 has also been highlighted as a mediator of lung fibrosis that acts via mechanisms that involve TGFβ signalling 153 . Thus, modulation of GAL1, GAL3 and/or GAL9 might control fibrotic processes by targeting fibroblasts, macrophages and ECs in several prominent pathological conditions, including IPF and NASH.

Galectin-targeted strategies
As our understanding of the role of galectins in cancer and fibrosis has improved, various chemical and biological agents that target GAL1, GAL3 and GAL9 have been developed. These agents include small-molecule chemical inhibitors, natural polysaccharides and their synthetic derivatives, peptides and peptidomimetics, and biological agents such as small interfering RNA (siRNA), aptamers, truncated galectins and neutralizing mAbs. In the next section, we describe both the advances and the difficulties in the development and clinical evaluation of current galectin-targeted agents.

Small-molecule carbohydrate inhibitors
The binding of galectins to β-galactosides such as LacNAc (1, Fig. 3) is determined by several distinct non-covalent interactions, including a hydrophilic hydrogen-bond network between the sugar hydroxyl groups and the conserved amino acids Arg, His, Asn and Glu in the galectin. Other interactions include hydrophobic CH-π bonds between the galactoside and a Trp-containing side chain in the galectin CRD 159 (Fig. 3a). The first rational approach to design galectin-antagonizing agents focused on small β-galactoside derivatives 160 . However, single carbohydrate-lectin interactions are typically low affinity, selectivity is not easily achieved, and carbohydrate molecules are cleared rapidly from the systemic circulation and are susceptible to degradation by glycosidases. Modifications at the C1 and C3 positions of monosaccharide β-galactose-based antagonists resulted in new molecules with improved interaction profiles within the galectin ligand-binding groove. These modified β-galactoside-based antagonists had enhanced affinity for galectins as determined by NMR and X-ray crystallo graphy [161][162][163] , but remained susceptible to hydrolytic degradation in vivo because of their labile O-glycosidic bonds.
Studies also documented the design of aminopyrimidinyl α-thiogalactoside (3), which, despite its comparatively low affinity for GAL3 (K d = 1.7 μM) compared with GB1107 (2), proved to be a more selective GAL3 inhibitor. Specifically, it showed >500-fold selectivity compared with GAL1 and a range of 34-to >500-fold selectivity compared with other galectins tested as determined by fluorescence polarization assays 170 (Fig. 3b). Furthermore, a patent assigned to Bristol Myers Squibb reporting α-thiogalactosides derived from GB1107 bearing additional chemical modifications, such as alkylation of HO-2 and

Review article
oxidation of the anomeric thiol group, was recently published 171 . Moreover, Idorsia Pharmaceuticals patented α-C-galactopyranosides bearing modifications on HO-2, HO-3 and the anomeric group 172 . Galecto developed GB1211, an α-thiogalactoside that is closely related to GB1107 with excellent affinity for GAL3 (K d = 23 nM) 173 . GB1211 was characterized as a highly selective, orally available GAL3 inhibitor, and its therapeutic potential was demonstrated in mouse models of CCl 4 -induced liver fibrosis and bleomycin-induced lung fibrosis 174 . Orally administered GB1211 was well tolerated in phase I clinical studies in healthy subjects (NCT03809052, Table 1). A phase I/II study of this compound in patients with confirmed NASH and liver fibrosis was withdrawn (NCT04607655 ,  Table 1). However, a phase I/II trial of GB1211 in participants with hepatic impairment (Child-Pugh class B and C) is ongoing (NCT05009680, Table 1). Galecto recently presented initial results from this phase I/II trial indicating that GB1211 is safe and well tolerated 175 , and early signs of clinical effect in patients with moderate and severe hepatic impairment were noted 176 . A phase I/II trial to assess the combination of GB1211 with the anti-PDL1 mAb atezolizumab in patients with non-small-cell lung cancer (NSCLC) has started recruitment (NCT05240131, Table 1). To achieve higher selectivity towards GAL3, interactions between galectins and alternative monosaccharides, including d-mannose, d-talose and d-gulose derivatives have been evaluated, although these compounds exhibited no significant functional activity in in vitro or in vivo assays [177][178][179][180][181][182] . However, a recent study reported that GAL9C and GAL9N CRDs demonstrated good affinity towards synthetic 3-deoxy-3-N-arylated-α-d-galactoside (4) and α-guloside (5) derivatives, respectively, compared with other galectins, as determined by fluorescence anisotropy assays 183 (Fig. 3c).
Lactose (Lac) and LacNAc disaccharides, the natural galectin ligands in glycoconjugates, have been evaluated as galectin inhibitors 37,184 . Substitution of LacNAc at O3′ with aromatic moieties has been explored extensively as a means to generate new and favourable π-arene interactions with the conserved Arg144 residue; this effort has resulted in the development of several effective GAL3 ligands, including a LacNAc derivative 163,185,186 (6, Fig. 3d). Alternatively, synthetic lactulose amines evaluated as galectin modulators were found to induce tumour cell apoptosis and inhibit homotypic cell aggregation and EC morphogenesis 187 . More recently, Kishor and colleagues 188 explored interactions of GAL1 and GAL3 with lactulose (7, Fig. 3e) by crystallography and proposed this disaccharide as another scaffold for the design of novel galectin inhibitors.
S-, C-and N-disaccharides have been investigated as stable galectin inhibitors. The synthetic non-reducing disaccharide thiodigalactoside (TDG) (8), a compound that binds to galectins with similar affinity to LacNAc but lacks labile glycosidic bonds, emerged as a valuable building block for the preparation of stable galectin ligands 189,190 (Fig. 3f). Similarly, previous work on galactose and lactose derivatives led to the symmetrical substitution of TDG at O3 and O3′ with aromatic esters, amides or triazole groups, which ultimately resulted in several galectin inhibitors with nanomolar affinities 164,[189][190][191] (Fig. 3f). Bis-[3-O-(3methoxybenzoyl)-β-d-galactopyranosyl] sulfane (named Td131_1) (9) and bis-(3-O-1-naphthoyl-β-d-galactopyranosyl) sulfane (10) (Fig. 3f) inhibited the migration of human prostate and NSCLC cell lines 189 . Td131_1 (9) was also tested in papillary thyroid cancer cells in vitro, and, despite the high concentrations required, it promoted cancer cell apoptosis and synergized with doxorubicin 192 . These successful results led to the evaluation of 3,3′-disubstituted TDG derivatives as GAL3 inhibitors in chronic inflammation and fibrosis 193 . For example, bis-(3-deoxy-3-(4-(3-fluorophenyl)-1H-1,2,3-triazol-1-yl)-β-d-galactopyranosyl) sulfane (11, Fig. 3f; also named TD139, 33DFTG, and later renamed GB0139 by Galecto), is a selective potent GAL1 and GAL3 inhibitor 145 with low affinity for GAL2, GAL4N, GAL4C, GAL8N and GAL9N 145 and low or no significant affinity for GAL7 (refs. 144,145 ). Inhibition of GAL3 with GB0139 (11) in a model of bleomycin-induced lung fibrosis resulted in decreased TGFβ receptor at the cell surface, together with suppressed β-catenin activation and attenuated severity of lung fibrosis 145 . This result was obtained when GB0139 was administered at much lower concentrations than pirfenidone, an anti-inflammatory drug approved for the treatment of IPF 194 . Bratteby et al. 195 recently developed surrogate positron emission tomography (PET) radiotracers for GB0139 and GB1107 (2) and described their pharmacokinetics. The 18 F-radiolabelled GB0139 surrogate was cleared much more rapidly than the GB1107-derived compound, suggesting that although systemic administration may be favourable for monosaccharide-based galectin inhibitors, the rapid excretion of disaccharide derivatives could limit their activity. A phase I/II clinical trial initiated in 2014 to assess the safety, pharmacokinetics and tolerability of GB0139 via a dry-powder inhaler in human volunteers and patients with IPF (NCT02257177) showed that this compound was safe and well tolerated. Furthermore, administration via this route inhibited GAL3 expression by bronchoalveolar lavage macrophages and decreased the concentration of plasma biomarkers associated with IPF progression 23 . A phase IIb clinical trial to investigate the safety and efficacy of GB0139 (11) in patients with IPF has completed enrolment (NCT03832946). As TDG-based ligands progressed towards clinical trials, further studies to improve selectivity highlighted that their substitution with bulky aromatic groups favours interactions with GAL3 over GAL1 (ref. 196 ). Consistent with these results, a coumarin-derived TDG (12) showed a 175-fold increase in selectivity over GAL1 with similar efficacy to GB0139 in the bleomycin-induced model of lung fibrosis 197 (Fig. 3f). Recently, tetrahydropyran-based thiodisaccharide mimics with a reduced number of H-bond donors to improve permeability and bioavailability were also described 198 .

Review article
a natural glycopeptide rich in Galβ(1,3)GalNAc structures isolated from the cod antifreeze glycoprotein binds to GAL3 with high affinity (K d 97 pM) and inhibits adhesion of PC-3 prostate cancer cells to ECs as well as GAL3-mediated T cell apoptosis 206 . Moreover, complex type N-glycans prepared from chicken eggs and coupled by click chemistry to human serum albumin (HSA) as scaffold led to multivalent GAL3 ligands 207 . More recently, the chemoenzymatic synthesis of neoglycoproteins based on coupling of C3-substituted LacdiNAc glycomimetics

Review article
to HSA (15, Fig. 3g) has been reported; these derivatives protected T lymphocytes against GAL3-induced apoptosis in vitro 208 .
Given that the functions of galectins vary between the intracellular and extracellular compartments 16,46 , Stegmayr et al. 209 investigated the intracellular activity and membrane permeability of selected smallmolecule GAL3 inhibitors including compounds GB1107 (2), GB0139 (11) and a 1H-1,2,3-triazol-1-yl TDG derivative 210 . Inhibitor uptake and intracellular potency correlated with the polar surface area of the compounds, which is a determinant for passive membrane permeability. Of the compounds investigated, GB1107 was the most potent, although GB0139 had increased intracellular activity following pre-incubation, suggesting that it can reach the cytosolic compartment given sufficient time to cross the plasma membrane 209 . Thus, small-molecule galectin inhibitors that can cross the cell membrane, such as GB0139, are important tools to discern extracellular and intracellular functions of galectins, and could potentially be used to target intracellular galectins.
Pectins are a class of complex polysaccharides found in the primary cell walls of plants and are composed of an anionic galacturonan backbone (a linear chain of α(1-4)-linked d-galacturonic acids) and neutral sugar side chains that contain primarily d-galactose, l-rhamnose and l-arabinose 215,216 . Underivatized pectins have shown beneficial health effects, primarily when introduced as dietary fibre to prevent colon cancer 217 ; however, the most promising anticancer effects were observed in response to MCPs. Treatment of pectins to obtain MCPs eliminates galacturonic acid esters from the homo galacturonan regions and exposes galactose-containing side chains, which have been proposed as major determinants for galectin binding. However, the affinity and selectivity of these compounds is still under debate, as little or no inhibitory activity was detected by fluorescence anisotropy 218 . Nevertheless, MCPs have been extensively studied for potential antitumour and antimetastatic activities in several cancer types 211,212,219,220 . For example, PectaSol and PectaSol-C (16) (Fig. 4) are commercial MCPs developed by EcoNugenics that have cytotoxic activity against murine and human prostate cancer cell lines via suppression of MAPK signalling and induction of caspase 3 cleavage 213 . PectaSol-C functions synergistically with paclitaxel to promote cytotoxicity in ovarian cancer cells by modulating GAL3-driven activation of signal transducer and activator of transcription 3 (STAT3) 221 . In 2003, a phase II pilot study 222 evaluated tolerability and efficacy of PectaSol-C in patients diagnosed with prostate cancer. Oral administration of PectaSol-C resulted in a statistically significant increase in the prostate-specific antigen (PSA) doubling time, potentially indicating a slowing of tumour progression. In a subsequent phase II clinical trial in patients with biochemical relapse of prostate cancer, oral administration of PectaSol-C was evaluated by measuring PSA kinetics as a potential marker for cancer progression (NCT01681823, Table 1). Results showed that patients treated with PectaSol-C exhibited no PSA progression or lengthening of PSA doubling time compared with historical data, potentially indicating slower disease progression 223

Review article
GCS-100 (17, Fig. 4), developed by La Jolla Pharmaceutical Company, was initially developed as a potential inhibitor of GAL3. GCS-100 has broad antitumour activity, including inducing apoptosis of multiple myeloma cells 212,[224][225][226] . Moreover, GCS-100 alone or in combination with BCL-2 homology domain 3 (BH3) mimetics induced apoptosis in acute myeloid leukaemia cells 227 . A phase II clinical trial (NCT00514696) evaluated intravenous administration of GCS-100 in patients with chronic lymphocytic leukaemia. The inhibitor showed excellent tolerability and led to partial remission in 25% of patients and >50% shrinkage of lymph node lesions in 16% of patients.
GR-MD-02 (18) (renamed belapectin by Galectin Therapeutics) (Fig. 4), a pectin-derived galactoarabino-rhamnogalacturonate substituted with predominantly β(1-4)-d-galactose and α(1-5)-l-arabinose side chains, has been proposed as a GAL3 inhibitor. The antitumour effects of belapectin were explored in immunocompetent cancer models in combination with an agonist anti-OX40 antibody. This drug combination promoted tumour regression and increased survival in mouse models by mitigating MDSC-driven immunosuppression, increasing CD8 + T cell recruitment and reducing the frequency of CD4 + FOXP3 + T reg cells 122 . This compound was also tested in two phase I clinical trials in combination with mAbs that inhibit the immune checkpoint proteins CTLA4 (ipilimumab) or PD1 (pembrolizumab) in patients diagnosed with melanoma (NCT02117362 and NCT02575404). A combination of belapectin and pembrolizumab administered intravenously to patients with advanced metastatic melanoma or head and neck squamous cell carcinoma (HNSCC) (NCT02575404) was well tolerated and generated fewer immune-mediated adverse events than anticipated, based on findings from pembrolizumab monotherapy. Moreover, belapectin and pembrolizumab treatment enhanced the activation of effector memory T cells and decreased the frequency of MDSCs in patients responding to this regimen (50% of patients with metastasic melanoma and 33% of patients with HNSCC) 228 .
GM-CT-01 (19), also known as Davanat (Galectin Therapeutics; Fig. 4), is a chemically modified galactomannan from the plant Cyamopsis tetragonoloba (guar gum) that has been shown to enhance the efficacy of the chemotherapeutic agent 5-fluorouracil (5-FU) in models of colon and breast cancer 229 . These results led to phase I (NCT00054977) and phase II (NCT00110721) clinical trials that demonstrated a lack of toxicity and a 46% increase in survival of patients with colorectal cancer who received this drug combination 214,230 . A comparison of GCS-100 (17) and Davanat (19) suggests that these two polysaccharides might have different mechanisms of action. Although the binding site for Pec-taSol or GCS-100 is still not clear 218 , Davanat binds to the F-face of the GAL3 β-sandwich and not to the canonical site for β/α-galactosides 231 (Fig. 4b).
Plant-derived polysaccharides have also been proposed as therapeutics for liver, kidney and lung fibrosis, mainly via mechanisms that involve inhibition of GAL3 (refs. 142,146 ). Inhibition of GAL3 with belapectin (18, Fig. 4) and Davanat (19, Fig. 4) was first evaluated in a toxin-induced model of liver fibrosis 232 . Intraperitoneal administration of these polysaccharides resulted in decreased collagen content, attenuated liver fibrosis, diminished cirrhosis and a reduced percentage of GAL3-expressing macrophages 232 . These two inhibitors were also tested in a murine model of NASH 233 . Intravenous administration of belapectin resulted in a substantial reduction in collagen deposition, hepatocellular damage, NASH activity and fibrosis -features that were associated with reduced markers of inflammation that included inducible nitric oxide synthase (iNOS) and CD36 + pro-inflammatory macrophages. By contrast, administration of Davanat had no effect 233 .
A phase I clinical trial of belapectin (18) in patients with NASH with advanced hepatic fibrosis revealed no toxicity and good tolerability 234 (NCT01899859, Table 1). Two phase II clinical trials evaluated the efficacy of this compound in liver fibrosis. In patients with NASH with advanced fibrosis (NCT02421094, Table 1), belapectin had no significant effects on levels of non-invasive biomarkers of liver inflammation or fibrosis over a 4-month period 234 . In liver fibrosis and resultant portal hypertension in patients with NASH cirrhosis 235 (NCT02462967, Table 1), belapectin had no impact on fibrosis or nonalcoholic fatty liver disease activity score. However, in a patient subgroup it showed a significant effect on portal pressure and prevented the development of oesophageal varices, which is an early sign of serious complications in patients with cirrhosis. This led to the development of a phase IIb/III trial designed to evaluate its safety and efficacy specifically in patients with NASH-associated cirrhosis for the prevention of oesophageal varices (NCT04365868, Table 1).
Other polysaccharides have also emerged as potential galectin inhibitors. For example, RN1 (20, Fig. 4), an arabinogalactan polysaccharide isolated from the flowers of the Chinese ginseng plant (Panax notoginseng), was proposed as a GAL3 inhibitor. This compound showed antitumoural activity in pancreatic ductal adenocarcinoma both in vitro and in vivo 236 . Other ginseng-derived pectins prevented GAL3-driven T cell apoptosis and reduced tumour growth in mouse models of sarcoma 237 .

Peptides and peptidomimetics
A peptide known as Anginex (βpep-25, 21, Fig. 5) that was designed based on the structure of anti-angiogenic proteins was later found to be a potential anti-galectin therapeutic 238,239 . Anginex is a 33-mer with an amphipathic β-sheet that initially lacked a specific molecular target. However, several studies revealed that this compound had potent anti-angiogenic activity via its capacity to block EC adhesion and migration, ultimately leading to inhibition of angiogenesis and tumour growth [240][241][242] . GAL1 was subsequently identified as the molecular target of Anginex using fluorescence microscopy, NMR and surface plasmon resonance (SPR) analysis, with a strong binding constant (K d 90 nM) (ref. 243 ). However, despite these initially promising results, Anginex was beset by several drawbacks that hampered its translation to clinical settings; these include its complex chemical synthesis, short half-life and poor stability. Aiming at a more affordable alternative than peptide synthesis, an artificial gene encoding Anginex was developed, offering a recombinant peptide with the same structure and slightly decreased activity 244 . To overcome its rapid clearance and sensitivity to endogenous endopeptidases, an adeno-associated virus vehicle delivery system was evaluated 245 . Using this delivery system, Anginex suppressed proliferation, migration and invasion of human umbilical vein endothelial cells (HUVECs) and inhibited angiogenesis and tumour growth in an ovarian cancer xenograft model. Finally, the Anginex derivative 6DBF7 (22, Fig. 5), a dibenzofuran (DBF)-based partial peptidomimetic, demonstrated improved stability 246 . 6DBF7 retained anti-angiogenic and tumour growth inhibitory activity in ovarian cancer mouse models and was even more effective in vivo than Anginex 246 . Dings et al. 247 designed and tested new optimized analogues of 6DBF7 (22) and introduced DB21 (23, Fig. 5), a derivative capable of increased inhibition of angiogenesis and tumour growth in melanoma, lung adenocarcinoma and ovarian cancer models. Both 6DBF7 and DB21 were identified as GAL1 inhibitors, although no data on their interactions with other members of the galectin family are available.

Review article
More recently, OTX008 (or 0118, also called PTX008; 24, Fig. 5) was developed as a non-peptidic calixarene-based Anginex topomimetic with GAL1 binding capacity 248 . OTX008 inhibited angiogenesis in both in vitro and mouse model systems 99,249 . It also synergized with the tyrosine kinase inhibitor sunitinib in ovarian carcinoma and glioblastoma xenograft models 250 . A phase I clinical trial sponsored by Oncoethix assessed the impact of OTX008 administered subcutaneously to patients with advanced solid tumours and documented its rapid absorption and urinary excretion (NCT01724320) 251 . More recently, Leung et al. 252 demonstrated that OTX008 combined with sorafenib reduced tumour growth and enhanced the therapeutic effects of sorafenib alone in experimental models of hepatocellular carcinoma.
OTX008 was also effective against thyroid cancer in in vitro and in vivo studies 253 . Other calixarene derivatives include the polycationic calixarene-based compound PTX013 (25), which displayed 50-fold enhanced activity over that of OTX008 in an experimental melanoma model 254 . Furthermore, new analogues showed increased cell growth inhibitory activity in HUVECs and MA148 ovarian cancer cells, when compared with the parent calixarene compound 24 (ref. 255 ). However, the quest for improved activity led to 'off-target' side effects: PTX013 was toxic in mice, and considering that, in vitro, PTX013 could be considered a cytotoxic drug more than a cytostatic agent (like parental PTX008), the authors postulated that GAL1 might not be the only target of this compound 254,256 . In this sense, PTX013 has recently been   . h, GAL1 binding site for OTX008 (24). Amino acid residues that interact with OTX008 are highlighted in orange (moderate interaction) or red (high interaction). NB, no binding observed; ND, not determined.

Review article
shown to bind to both GAL3 and GAL1; this was suggested as a potential mechanism underlying its increased activity 257 . The development of PTX013 is an important example of how structural modifications aimed at improving biological activity can result in increased toxicity via potentially different mechanisms of action 254 . An alternative approach to designing galectin-targeting peptides was developed by the Deutscher group 258 , which generated two small synthetic 15-mer peptides, G3-A9 (26, Fig. 5) and G3-C12 (27, Fig. 5), which were identified by combinatorial bacteriophage display. These peptides were selective for GAL3, but not for GAL1 or GAL4. They blocked the interaction of GAL3 with the Tf antigen and inhibited the adhesion of MDA-MD-435 human breast carcinoma cells to ECs in vitro. Peptide G3-C12 (27) is antimetastatic via its capacity to modulate GAL3 activity 258 ; its administration resulted in a 72% reduction in MDA-MB-231 cell adhesion to vasculature in nude mice. However, despite its therapeutic potential, in vivo studies showed that G3-C12 accumulated in the kidneys 259 .
To overcome this limitation, control its renal clearance and improve its therapeutic efficacy, G3-C12 (27, Fig. 5) was conjugated to an N-(2-hydroxypropyl)methacrylamide polymeric carrier. This substitution increased its molecular size and generated a larger GAL3targeted copolymer that was proposed as an anticancer agent 260,261 . Conjugation of HMPA-G3-C12 copolymer to 5-FU led to increased cytotoxicity, tumour cell apoptosis and inhibition of migration in vitro compared with 5-FU, as well as greater reductions in tumour volume in a PC-3 prostate cancer mouse model 262 . Micellar nanoparticles assembled by poly(oligo(ethylene glycol)) monomethyl ether methacrylate (POEGMA) and poly(ε-caprolactone) (PCL) copolymers conjugated to G3-C12 peptide were designed to deliver G3-C12 (27, Fig. 5) to the tumour microenvironment. This copolymer formed nanoparticles that were loaded with the anticancer drug bufalin (BUF) and tested as a drug delivery system for castration-resistant prostate cancer (CRPC). Biodegradable BUF-G3-C12 nanoparticles exhibited controlled drug release and enhanced tumour reduction in mice bearing DU145 prostate cancer cells 263 .

Biological agents
To avoid 'off-target' effects and achieve higher selectivity, efforts are underway to develop biological agents that target individual members of the galectin family, including galectin-specific neutralizing mAbs, decoys, aptamers and RNA silencing strategies using siRNA 9,98,100,264-267 . Both RNA interference (RNAi) and CRISPR technologies have been used to inhibit the expression and function of specific galectins in several experimental models 268 . GAL1 silencing with RNA technologies led to dramatic antitumour effects in several cancer models 9,98,114,125,126,128,256 by unleashing antitumour immunity and suppressing aberrant vascularization 9,98,266 . More recently, administration of a GAL1-targeted DNA aptamer resulted in antitumour effects via its capacity to restore immune function in vivo 267 . Despite promising results in preclinical models, more investigations into the mechanisms that underlie nucleic acid-based targeting strategies will be required to translate these findings into clinically important therapies.
Over the past two decades, mAb-based therapies that target immune checkpoint or pro-angiogenic pathways have generated unprecedented success in treating a broad range of malignancies 269 . mAb-based therapeutics directed against galectins could be an important approach, as these agents are target specific and well tolerated in humans 270 . To date, several GAL1-neutralizing mAbs have been developed, characterized and functionally evaluated in in vitro and in vivo models 9,98,100,271,272 . For example, anti-GAL1 mAbs inhibit GAL1mediated apoptosis of Epstein-Barr virus (EBV)-specific CD8 + T cells, suggesting that they could prevent EBV-induced immune evasion in post-transplant lymphoproliferative disorders (PTLDs) 271 . Moreover, these mAbs restrained tumour growth in models of Kaposi's sarcoma, lung adenocarcinoma, melanoma and T cell lymphoma 9,98 . Because GAL1 is known to activate VEGFR2 signalling 9 , the ability of anti-GAL1 mAb to prevent resistance to VEGF blockade was assessed in various models. Blockade of GAL1-N-glycan interactions using a functional blocking mAb not only circumvented resistance to anti-VEGF but also normalized aberrant tumour vasculature and promoted an increased influx of immune cells to the tumour bed 9 . Given these promising findings, additional mAbs specific for GAL1 have been designed and found to have therapeutic potential 100,272 . Interestingly, mAb-mediated GAL1 blockade enhanced the immunostimulatory effect of anti-PD1 therapy in HNSCC by preventing T cell exclusion from tumour sites 92 . Moreover, a recombinant vaccine protein consisting of bacterial sequences fused to GAL1 was recently described to generate high anti-GAL1 antibody levels in experimental models of melanoma, associated with increased cytotoxic T cell infiltration and decreased tumour burden 273 .
Based on the oncogenic roles of GAL3 triggered upon interaction with N-glycans on MUC16 (ref. 274 ), Stasenko et al. 275 evaluated the therapeutic potential of anti-GAL3 mAbs in high-grade serous ovarian cancer and other MUC16/CA-125-expressing malignancies. The GAL3blocking mAb 14D11 recognizes both human and mouse GAL3, but not other galectin family members. Administration of this agent extended overall survival in tumour-bearing mice and delayed the occurrence of lung metastasis in breast cancer models 275 . Recently, anti-GAL3 mAbs developed by the company TrueBinding were patented for the treatment of liver fibrosis, pulmonary fibrosis, NASH and various neoplastic conditions 276 .
Furthermore, given the tolerogenic activities of GAL9 in Dectin-1-expressing macrophages 124 , a fully human IgG4 mAb to GAL9 known as LYT200 was developed by PureTech Health, and was the first antigalectin mAb to reach clinical evaluation. This agent is in phase I/II clinical trials either alone or in combination with chemotherapy (gemcitabine/nab-paclitaxel) or anti-PD1 mAbs as a potential treatment for metastatic solid tumours (NCT04666688, Table 1). In 2021, the FDA designated LYT200 an orphan drug for the treatment of pancreatic cancer. Additionally, anti-GAL9-based therapies (including LYT200) exhibit cytotoxic effects in models of human T cell acute lymphoblastic leukaemia (T-ALL) 277 . Thus, GAL1-, GAL3-or GAL9-blocking mAbs, either alone or in combination, represent an array of multifunctional and selective strategies that can reprogramme the tumour, immune and vascular landscapes in a broad range of malignancies.
Another strategy for GAL3 blockade was the development of an N-terminally truncated form of this lectin (GAL3C); the N-terminal fragment is not involved in carbohydrate recognition but is essential for cooperative binding and cross-linking. As a consequence, GAL3C functions as a competitive inhibitor of GAL3 by preventing homophilic cross-linking following its interactions with glycan structures. Recombinant GAL3C has advantages such as a relatively simple and controlled production process and high selectivity compared with other synthetic inhibitors and natural polysaccharides 278 . GAL3C exhibited antitumour and antimetastatic activities in models of breast 279 and ovarian 278 cancer via mechanisms involving CD44 inactivation and inhibition of the effects of MUC1, including abrogation of AKT and nuclear factor-κB (NF-κB) downstream signalling pathways. Furthermore, intravenous administration of GAL3C in a multiple myeloma model resulted in

Review article
diminished tumour growth and increased the therapeutic impact of the proteasome inhibitor bortezomib 280 . However, a major limitation of GAL3C is its rapid clearance; hence, efforts are underway to explore the effects of GAL3C conjugation with other molecules to increase its molecular weight and stability. In this regard, the fusion of GAL1 with the Fc region of human IgG1 (GAL1hFc) resulted in enhanced stability and lifespan of this lectin accompanied by more pronounced immunoregulatory activity 134 . Thus, structural modifications might improve galectin-mediated therapeutic efficacy, mitigate 'off-target' effects and amplify clinically relevant antagonistic or agonistic activities.

Future potential of galectin inhibitors
Several challenges need to be addressed to accelerate the progress of galectin inhibitors into clinical settings. Some challenges are specific to particular therapeutic modalities; for example, compared with peptides, pectins or mAbs, small-molecule inhibitors such as galactosides or lactosides are more challenging to synthesize, and the development of new synthetic methods for these types of compound is currently an active area of research 281 . Other challenges are common to some or all of the modalities, including thorough understanding of their selectivity, biodistribution and safety profiles, and dissection of specific mechanisms of action. Below, we compare the advantages and limitations of various galectin-targeted therapeutic modalities and consider their future clinical potential.

Understanding selectivity
A key challenge in the design of galectin inhibitors is to avoid 'off-target' effects by preventing binding to other cellular proteins, including other family members. Although GAL1, GAL3 and GAL9 seem to share a common pro-fibrotic and pro-tumoural function, they might have contrasting roles in other pathological conditions. Moreover, given their structural similarities and wide tissue distribution, selective targeting of specific galectins could also become relevant in fibrotic or neoplastic diseases. To achieve selective targeting, it might be necessary to exploit the differences in galectin expression pattern and biological activity as well as differences in their ligand-binding specificity, capacity to cross-link individual glycosylated receptors and extracellular or intracellular distribution. Moreover, biochemical factors (stability, redox status, pH and multimerization) that govern their glycan-dependent or independent interactions should also be considered [282][283][284] .
The extensive investigations into higher affinity and more selective small-molecule galectin inhibitors has led to an improved understanding of the crucial interactions that underlie their inhibitory activity and enhanced selectivity 179 . The use of GB0139 (11) for IPF and GB1211 for NASH laid the foundations for translational studies focused on smallmolecule galectin inhibitors. GB0139 was first described as a potent inhibitor of GAL3 but also exhibited affinity for GAL1 (ref. 144 ). This finding highlighted the potential use of these compounds for therapeutic targeting of multiple galectins, including those with pro-fibrotic activity.
Regarding peptidomimetics, most of the beneficial effects of Anginex have been associated with GAL1 inhibition, although affinity for GAL2, GAL7, GAL8N and GAL9N has also been demonstrated 285 , suggesting that Anginex and related compounds have limited selectivity. Additional work is needed to explore the selectivity of peptide-based inhibitors towards different galectins and potential 'off-target' effects.
Modified natural complex polysaccharides such as MCPs or galactomannans are well tolerated in humans with few adverse effects. However, they vary substantially in molecular weight and degree of esterification, and their non-synthetic nature limits the thorough structural characterization required to understand their selectivity. Of particular note, the affinity of galectins for pectins and galactomannans has been questioned 218 . GCS-100 has been proposed as a GAL3 inhibitor, although its selectivity with respect to other galectins remains unknown. Further biochemical characterization of galectinpolysaccharide interactions is required, as well as in vivo studies on the biological activities of these compounds in galectin-null mice, as the possibility that polysaccharides could interact with other key mediators of disease pathogenesis has not been ruled out 218 .
Despite its enhanced selectivity when compared with other inhibitors, truncated GAL3C has not yet reached clinical trials. The specificity of anti-GAL1 and anti-GAL3 mAbs has been demonstrated in in vitro assays 9,35,272,275 , although further studies performed in vivo are needed to examine potential 'off-target' effects.
For these highly specific biological agents, it is crucial to recognize that neoplastic diseases present a case-by-case challenge; in some cases, selectivity for a specific galectin might not be crucial because two or more galectins exert similar biological activities (for example, immuno suppression driven by GAL1, GAL3 and/or GAL9). In these cases, overall galectin activities could be blocked using a pan-galectin inhibitor. The varied expression levels, diverse cell sources and possible opposing functions of galectins based on extracellular or intracellular localization should be examined in individual cancers. Recent advances in human organoids and humanized animal models, together with the development of single-cell technologies [286][287][288] and machine learning algorithms, will be invaluable to discern the source and cell-specific function of these lectins in the tumour microenvironment. Furthermore, the exquisite regulation by glycan epitope density on specific receptors as well as the multivalent nature of these interactions should also be considered 58 . In this respect, multivalent glycodendrimers, sugar calixarenes and nanotechnology-based developments could be useful to understand the natural clustering and presentation of ligands on the target cell surface, providing crucial mechanistic insights into their multivalency. A low-affinity inhibitor might be presented in clusters, which could increase its potency. As an example, lactose calixarenes were described as selective GAL4 inhibitors in vitro, with 300fold improvement in binding 289 . However, these large, polar and labile molecules could present a difficult challenge owing to their ADME(T) properties.
Overall, a deeper understanding of the multifunctionality and disease-specific expression profiles of the various galectins, improved characterization of the cellular glycome and further identification of galectin binding partners and receptors will be essential for the design of novel galectin inhibitors with various degrees of selectivity.

Identifying mechanisms of action
The mechanisms of action of currently available galectin inhibitors differ considerably from one another, and, in some cases, remain a matter of ongoing debate. Although plant-derived polysaccharides such as MCPs or galactomannans have reached clinical trials, their specific targets and/or mechanisms of action remain unclear 218,[290][291][292] . The apparent contradiction between the limited interactions with galectins and the health-promoting effects of these polysaccharides was evaluated by Miller et al. 231,293,294 , who suggested that the biological effects do not involve interactions with canonical galectin CRDs. Indeed, Stegmayr et al. 218 have proposed that Davanat inactivates GAL1 by an oxidative mechanism. Likewise, the negative charge of pectins suggests that their Review article therapeutic effects could be dependent on charge-based interactions instead of specific intermolecular affinities. Other studies of pectinderived polysaccharides focused on their direct cytotoxic effects involving apoptosis and autophagy 213,295 . Moreover, it is not clear whether these high molecular weight polysaccharides target the intracellular functions of galectins via internalization (as observed in the case of dextrans), followed by selective distribution to various subcellular compartments 296 . Additionally, these macromolecules might function by sequestering galectins in the extracellular space, thereby modifying their distribution between the intracellular and extracellular compartments. Although this is an attractive hypothesis, intracellular levels of GAL3 did not change upon the addition of MCPs 218 . Additional studies focused on the mechanisms that underlie the uptake, function and target engagement of plant-derived polysaccharides will be necessary to improve our understanding of their therapeutic potential.
Interestingly, regarding peptidomimetics, the DBF-based analogues 6DBF7 and DB21 inhibited tumour growth in mice to an even greater extent than Anginex, possibly owing to their increased bioavailability

Review article
and/or potential differences in mechanisms of action 242,247 . Whereas Anginex paradoxically enhances GAL1 binding to several oligosaccharides and glycoproteins 285 , the peptidomimetics 6DBF7 and DB21 behave as non-competitive allosteric inhibitors of GAL1 that reduce its affinity for lactose 247 . Topomimetics OTX008 and PTX013 also function as allosteric inhibitors of GAL1, albeit via an interaction site that differs from the one used by Anginex derivatives (Fig. 5). Like 6DBF7 and DB21, and in contrast to parental Anginex, these compounds decrease the affinity of the GAL1 CRD for lactose 254 . Collectively, these results suggest that structural modification of these peptides results in diversification of their mechanisms of action. Interestingly, using a recombinant adeno-associated virus to deliver Anginex revealed that this compound inhibited VEGF expression and the AKT, JNK and NF-κB downstream signalling pathways 245 . These GAL1-independent effects, together with the ability of Anginex to disrupt EC membranes via electrostatic interactions 297 , point to alternative modes of action. Additional in vivo studies in GAL1-deficient mice might help to unravel these effects.

Clinical outlook
The development of small-molecule galectin inhibitors has progressed towards high-affinity agents, surmounting limitations such as low stability and lifespan and leading to two pioneering compounds currently under clinical evaluation: the TDG derivative GB0139 for IPF, and the α-thiogalactoside GB1211 for NASH, liver fibrosis and NSCLC. Given the results obtained in clinical trials, their therapeutic potential is promising in both fibrosis and cancer. For example, a phase II trial to evaluate the safety and efficacy of GB1211 in combination with pembrolizumab in HNSCC and melanoma is expected to begin in 2023.
Regarding polysaccharides, clinical evaluation of GCS-100 and galactomannan Davanat has not been resumed, whereas PectaSol-C is commercially available as a dietary supplement for health and oncology support. Future studies will be crucial to unveil the full potential of the galactoarabino-rhamnogalacturonate belapectin, which is currently in phase IIb/III clinical trials for the prevention of eosophageal varices in NASH, and in the early stages of evaluation for melanoma, NSCLC and HNSCC in combination with pembrolizumab.
Among peptide and peptidomimetics, only OTX008 has been clinically evaluated, and this was more than a decade ago. Although the results of this phase I study documented its safety and low toxicity, no further advances have been described. Despite considerable efforts, our understanding of the mechanistic basis of the antitumour effects of peptide and peptidomimetic drugs beyond their specific interactions with galectins remains incomplete 252,253 .
Anti-galectin mAbs have been introduced more recently to reinforce antitumour immunity, reprogramme pathological angiogenesis and restrain tumour metastasis 9,98,124,272 . This concept is based on findings that suggest that anti-galectin mAbs have multifunctional roles as cancer therapeutics and that their use helps to avoid the accumulated toxicity associated with administration of multiple mAbs. The anti-GAL9 mAb LYT200 is under clinical assessment and is an example of the emerging strategy of targeting glycosylation-related immune checkpoints in cancer 14,298 . The use of intrabodies designed to target the intracellular functions of galectins, or nanobodies 299 capable of reaching areas within tumours and fibrotic foci, might be an important breakthrough in the coming years. Moreover, functionalized nanoparticles containing galectin-specific siRNA 265 , aptamers 267 and scFv fragments 300 might be alternative agents to target galectin-glycan interactions.
Finally, results obtained in clinical trials for NASH and IPF should facilitate the clinical translation of galectin blockers designed to treat other fibrotic conditions. For example, GAL3 has a crucial role in cardiac fibrosis leading to heart failure 301 , and clinical trials designed to evaluate PectaSol-C in patients with high blood pressure have been completed (NCT01960946) 302 . Moreover, inhibition of GAL3 with GB0139 (11) resulted in diminished VEGF-driven angiogenesis in a mouse model of corneal fibrosis 303 . In this regard, GAL1 has been proposed as a relevant target in renal fibrosis 304 , atherosclerosis and abdominal aortic aneurysms 34 , highlighting novel therapeutic opportunities for its inhibitors. Additionally, recognition of the central role of galectins in the pathogenesis of autoimmune inflammation 16 and neurodegeneration 30,31 has led to the design of galectin-based agonists and antagonists to treat these chronic inflammatory conditions. In this regard, TB006, an anti-GAL3 mAb developed by TrueBinding, is in a phase II clinical trial for treatment of acute ischaemic stroke (NCT05156827) and a phase I/II trial for Alzheimer disease (NCT05074498).
In conclusion, the rapid pace of advances in glycobiology, the promising performance of some galectin inhibitors in preclinical studies and clinical trials, and the urgent need for alternative therapies for cancer and fibrotic diseases should help to accelerate the translation of anti-galectin therapeutics into clinical practice.