Discovery and exploration of monosaccharide linked dimers of galectin-3 inhibitors to target fibrosis

Galectin proteins have been targets of interest in numerous therapeutic areas for some time. Galectin-3 has been identified as a target of particular interest because of its unique structural characteristics and physiological profile. Recent literature indicates galectin-3 inhibition can decrease myofibroblast activation and procollagen expression with the potential to affect the progression of fibrosis in the lung, liver and kidney. Potential π-stacking interactions between one monosaccharide ligand bound to the carbohydrate recognition domain of a galectin-3 protein and a second ligand bound to a different galectin-3 protein molecule were observed in the extended crystalline lattice in an X-ray crystal structure obtained while studying the monosaccharide structure-activity relationship. The direct interaction between the ligands suggests a potential for dimeric galectin-3 inhibitors which bind to two galectin-3 molecules simultaneously. This work describes the exploration of dimers designed to further probe these observations and explore the potential of binding to dimeric or higher order multimeric galectin-3 assemblies.


Introduction
Nonalcoholic fatty liver disease (NAFLD) has become a global concern as rates of obesity, type II diabetes and related metabolic diseases have increased and are inextricably linked to onset of the disease. Worldwide estimates in 2013 put the prevalence of NAFLD as high as one billion individuals currently living with the ailment [1]. In countries with a high prevalence of obesity and related disease, as many as a third of the population may be affected by NAFLD [1,2]. Further progression of liver disease in NAFLD patients can lead to the development of nonalcoholic steatohepatitis (NASH) which is more concerning as it is one of the principal causes of cirrhosis in US adults and can ultimately progress to hepatocellular carcinoma requiring a liver transplantation or result in liver-related mortality [2][3][4]. Because progression of NAFLD has thus far been difficult to predict, fibrosis has been the most impartial indicator of liver damage that will lead to severe disease [5,6].
Galectins are proteins that bind to conserved β−galactoside binding sites and have been studied for their numerous physiological functions in both the intra-and extra-cellular space [7][8][9][10][11]. Galectin-3 (Gal-3) has unique structure and function among the galectin family of proteins [12]. It is the only known chimera-type protein in the galectin family that is composed of an N-terminal non-lectin domain linked to a C-terminal carbohydrate recognition domain (CRD) [12][13][14][15][16][17]. Although Gal-3 is found predominantly as a monomer in solution, it can oligomerize to form disorganized, pentameric cross-linked complexes when interacting with multivalent carbohydrates [12,16]. It has been postulated that the ability for Gal-3 to organize in oligomeric forms and its ability to organize less structured cross-linked lattices compared to other galectins may account for differences in biological activity compared to other galectin proteins [12].
Extensive efforts have advanced the understanding of the roles that Gal-3 plays in the pathogenesis of various diseases [18][19][20][21]. Gal-3 is involved in numerous cellular and physiological processes such as apoptosis, cell adhesion and migration, angiogenesis, and inflammation [22]. Our group was especially interested in Gal-3's function in the regulation of inflammation and its therapeutic potential to affect the onset and progression of fibrotic diseases. Numerous studies indicate that Gal-3 is involved in myofibroblast activation and procollagen expression. Increased Gal-3 expression is associated with the progression of fibrosis in multiple organs including liver, lung, kidneys and the heart [22,23]. In the liver, Gal-3 expression is increased in hepatic stellate cells and human tissues of cirrhosis patients [24]. In preclinical models, Gal-3 inhibitors demonstrate therapeutic efficacies in acute CCl 4 liver injury models as well as thioacetamide induced liver fibrosis models in vivo [24][25][26][27].
Several research groups have advanced molecules into clinical trials for the treatment of fibrosis. TD139 (GB0139) is a thiodigalactoside being studied in phase IIb as an inhaled treatment of idiopathic pulmonary fibrosis [28]. TD139 was shown to have high binding affinity for Gal-3 in the low nanomolar range. However, this compound lacked oral bioavailability because of its large size and high polarity [29]. Further exploration led to the discovery of monogalactoside GB1211, a smaller and less polar low-nanomolar human Gal-3 inhibitor, that has improved oral bioavailability and is currently being examined in phase IIa clinical studies [29]. Together, publications outlining target engagement, predictive in vivo models, and advancement of compounds in the clinic make Gal-3 an attractive therapeutic target.

Results and discussion
During our effort to optimize the monosaccharide based Gal-3 series, crystal structure data aimed at guiding the structure-activity relationship (SAR) development provided several interesting observations [30]. An X-ray structure of 1 bound to the mouse galectin 3 protein (mGal-3) showed an apparent π−stacking interaction of the benzothiazole group extending from the anomeric 1,2,4-triazole of 1 with a benzothiazole on a second molecule of 1 (Fig. 1). Each of the two molecules of 1 were bound to a different mGal-3 protein molecule by key carbohydrate binding interactions in the CRD [30,31], while the two ligand molecules aligned in close proximity to one another in the extended crystalline lattice. The intermolecular π−stacking interactions between the ligand benzothiazoles may contribute to the stability of this particular crystalline form of the complex. Inter-ligand close contacts were also observed in the extended lattice of the X-ray crystal structure of human Gal-3 CRD bound to TD-139 (PDB ID: 7CXA) [32]. In another X-ray structure obtained from 2 bound to mGal-3 in the CRD, the close proximity of a carboxylic acid extending from the C2 position of the galactose core to the C2 carboxylic acid extension of a second molecule of 2 bound to a separate mGal-3 protein CRD was observed. The distance between oxygen atoms in the two carboxylic acid moieties was < 6 Å, suggesting the possibility of linking the two monosaccharide molecules as a bis-amide with a diamine of appropriate size. These observations led us to explore the possibility of developing a dimer that could bind to multiple Gal-3 proteins with the potential to further enhance binding affinity which is reported in the current paper.
Compounds were tested in both human and mouse Gal-3 (hGal-3 and mGal-3) homogeneous time-resolved fluorescence (HTRF) binding assays using biotin-asialofetuin (B-ASF) as the ligand. Many of the compounds were also tested in a second HTRF assay using a fluorescein isothiocyanate (FITC) labeled small molecule substrate [33] as the ligand (F-ligand). Both assays could be used as tools to effectively rank order the binding affinities of the molecules being tested for inhibition of Gal-3. Since we intended to use in vivo mouse models for PK/PD and efficacy studies, mGal-3 measurements became a focus for improvement with the monomeric series of compounds. In the case of compounds containing the trifluorophenyl substituents mG Ga al-3 3 / Co ompoun nd 1 1 2 m mGal--3 / / Com mpo oun nd Fig. 1 The X-ray crystal structure of mouse Gal-3 bound to 1 is shown on the left (PDB ID: 8ILU). Green ribbon represents one Gal-3 protein bound to green inhibitor molecule 1 while orange ribbon representing a second Gal-3 protein is in close proximity as it is bound to a second molecule of 1 in orange. The two benzothiazole moieties of 1 appear to associate through π-stacking (blue dashed lines) and the interaction may contribute to the stability of this crystalline form. At the right, an X-ray crystal structure (PDB ID: 8ILU) shows a distance of 5.8 Å between a carboxylic acid extension from the C2 position of one molecule of 2 to the carboxylic acid extension from the C2 position of a second molecule of 2 with an unobstructed path between the two moieties. Each molecule of 2 is bound to a distinct Gal-3 protein represented by the green and orange ribbons. Images created with the PyMOL Molecular Graphics System, Version 2.4 Schrödinger, LLC extending from the 1,2,3-triazole, mGal-3 potency was substantially lower than hGal-3 (19-fold for 1 and 15-fold for 2).
Several attachment points joining monosaccharide subunits were explored as possibilities to afford potent dimers based on the observed interactions in Fig. 1. Previously reported, monosaccharides containing 1,2,3-triazoles in the β conformation at the anomeric position showed moderate hGal-3 potency of 256-306 nM in the HTRF assay [30], so an attachment was made between the 4 position of the anomeric 1,2,3-triazole spanning 3 or 5 carbons between the monosaccharide subunits (3 and 4). Interestingly, these compounds showed an hGal-3 IC 50 > 4 μM and mGal-3 IC 50 > 30 μM in the HTRF assays. The dramatic decrease of the Gal-3 IC 50 value can likely be explained by the disruption of an important interaction at Gly182 of the protein with the halogen attached to a phenyl substituent projecting from the 1,2,3-triazole previously described by Liu et al. [30]. The benzothiazole π−stacking interaction observed with 1 likely does not disrupt the Gly182 interaction in the same manner since X-ray crystal structure data shows a similar interaction to the halogen bond can be made between the sulfur atom of the benzothiazole and the Gly182 of the Gal-3 protein [31].
Using a different approach to link the two monosaccharide subunits, an extension from the 1,2,3-triazole in one monosaccharide was linked to a C-2 alcohol of another monomer with an acetamide and a carbon spacer to provide 5. With this linkage, the hGal-3 and mGal-3 potencies were only several fold lower than that of the monosaccharide which had an hGal-3 IC 50 = 19 nM and mGal-3 IC 50 = 126 nM in the HTRF assay [30]. The structure of 5 allows the monosaccharide subunit with the extension from C-2 to interact with the Gly182 while maintaining the ability to halogen bond with the protein, while the second monosaccharide subunit is unavailable to interact with Gly182 since there is no longer an extension from the anomeric carbon able to access the same pocket. The observed potency, comparable to the single monosaccharide subunit, is consistent with this hypothesis.
To further test the effect of the linkages between the two subunits, additional connections between monosaccharides focused on side chains extending from the C-2 alcohol of the carbohydrate core which could allow for the halogen bond to Gly182 to be retained in each of the subunits. When comparing monosaccharide 6 to a symmetrical dimer made up of two monosaccharides linked through an acetamide spacer extending from the C-2 alcohol (examples 7-10, Table 1), the hGal-3 IC 50 was improved in each instance compared to the monosaccharide. The largest improvement in hGal-3 potency was 55x in the HTRF assay for 10 with a 5 carbon linker. Furthermore, 10 showed 7-13x improvement in the mGal-3 HTRF and F-ligand assays respectively compared to the monosaccharide. Since modifications to the phenyl substituents on the C-3 1,2,3-triazoles were reported to improve Gal-3 potency in the monosaccharide subunits one of the more potent substitution patterns, the phenyl F,Cl,F, was incorporated into dimers with acetamide linkers with 2, 4, 5, 6 and 7 carbon unit spacers. The 2, 4 and 5 carbon linkers (11)(12)(13) showed similar potency in the hGal-3 HTRF and F-ligand as the F,F,F-substituted analogs.
Linkers with 6 and 7 carbon spacers between the acetamides (14 and 15) showed lower Gal-3 inhibition in both the human and mouse assays than the shorter linkers of 11-13 providing the optimal length of the spacers between the acetamides of 2 to 5 carbons. Consistent with the reported monosaccharide SAR, The aryl substitution pattern showed a modest improvement in the mGal-3 potency for linkers of 2, 4 and 5 carbons (4-5 fold) in the fluorescence assay. Incorporating the F,Cl,F substitution pattern on the C3 triazole phenyl group appendage in combination with the CF 3 on R 2 of the 1,2,4-triazole, examples 16-18 provided another modest improvement to the in vitro mGal-3 IC 50 . Compounds 16-18 all had hGal-3 IC 50 of < 1 nM in the HTRF assay and importantly provided a 41-94 fold improvement in the mGal-3 HTRF assay and 23-74 fold improvement in the F-ligand assay when compared to the monosaccharide 6.
Because 10 showed a promising in vitro inhibition profile for both human and mouse, it was further profiled for advancement to in vivo studies. The in vitro permeability measured in the parallel artificial membrane permeability assay (PAMPA), a model for passive diffusion used to predict absorption [34], was 299 nm/sec at pH 5.5 and 376 nm/sec at pH 7.4, surprisingly high for a large, polar compound with a calculated topological polar surface area (tPSA) of 288 [35,36]. In vitro metabolic stability was measured by incubating 10 in human, rat, and mouse liver microsomes and measuring the % remaining after a 10 minute incubation period. Metabolism was generally low ranging from 84-92% across species. Mouse protein binding for 10 was 99.2% bound and human protein binding was 98.6% bound. With encouraging in vitro results, 10 was advanced to in vivo mouse PK studies. The compound was dosed at 2 mg/kg IV and 10 mg/kg PO and concentrations of the compound at multiple timepoints was acquired over a 24 h period. While the parent monosaccharide had an AUC total of 4.1 μM*h with a single dose IV and 4.0 μM*h with a single dose administered orally, only a small fraction of the dimer 10 was orally bioavailable (0.02 μM*h, PO, F% = 0.3) compared to the monosaccharide 6 which had F% = 20. With an aim to improve the oral bioavailability, alternative modifications to the linker were examined ( Table 2). The acetamide linkages of 7-18 were substituted for ether linkages of similar length in 19-20 which lowered the tPSA from 288 to 230 Å 2 . Although the distance between the two linked monosaccharide units was similar with the all carbon spacers in 19-20, HTRF assays for both the hGal-3 and mGal-3 showed a significant decrease in potency. When the linker was changed to a C-2 carbamate in 21 and 22, in vitro hGal-3 potency was somewhat restored with the 4 carbon linker between the carbamates having an hGal-3 IC 50 = 19 nM and the 3 carbon linker having an hGal-3 IC 50 = 110 nM. However, in both carbamate examples the mGal-3 IC 50 decreased to nearly 3 μM in the HTRF assay, a substantial loss of potency when compared to the original acetamide linkers. Finally, a tertiary amide 23 was explored which removed the H-bond donor characteristic of the initial secondary amides, however this changes only modestly decreased the tPSA to 270 Å 2 . Interestingly, in vitro hGal-3 and mGal-3 IC 50 was comparable to the secondary amides 11 with the same linker length. Unfortunately, attempts to measure the in vitro permeability using the PAMPA assay were not successful so 23 was not progressed further Table 3. To follow on this result, two cyclic tertiary amides were explored as linkers, the 1,4-diazepane 24 and the piperizine 25. Both compounds showed tPSAs of 270 Å 2 , and IC 50 values in the mGal-3 and hGal-3 F-ligand assay were comparable to the secondary amides of similar length. In addition, 24 and 25 showed acceptable metabolic stability in the 10 min incubation assay. However, attempts to measure in vitro permeability using the PAMPA assay were unsuccessful for both compounds, potentially because of limitations in aqueous solubility. The piperazine analog 25 was advanced into an in vivo mouse PK study to explore its potential as an orally dosed inhibitor of Gal-3. The compound was dosed at 10 mg/kg PO and 10 mg/kg SQ and concentrations of the compound at multiple timepoints were acquired over a 24 h period. When dosed orally, AUC total was only 0.035 μM*h, which was comparable to 10. The previous in vivo PO results of 10 in addition to a high tPSA and the difficulty measuring the in vitro permeability of 25 led to an alternative attempt at SQ dosing, which did improve the AUC total to 2.23 μM*h.
While exploring opportunities to link the two monosaccharide subunits together, attempts to further understand the binding pocket using binding models developed from X-ray crystal structures of monomeric subunits were explored, however attaining an X-ray crystal structure of one of the dimers was exceedingly arduous. After multiple attempts to crystallize the dimers in Table 1 came up empty, an X-ray crystal structure of 25 bound to the hGal-3 CRD was finally determined (Fig. 2). The X-ray crystal structure data clearly showed the two monosaccharide subunits interacting with two hGal-3 proteins in close proximity. One hGal-3 protein is shown in orange, while a second hGal-3 protein is shown in green and the dimeric ligand has organized in a symmetric, yet opposite orientation reaching into the binding pockets of each protein. When examining the binding of 25 to the hGal-3 protein closer (Fig. 3), it becomes apparent that the core CRD interactions of the saccharide are intact as well as the key interaction previously described with the anomeric substituent extending to reach Gly182 [30]. Interestingly, there is a water molecule that bridges the carbonyl of the acetamide linker, a ligand 1,2,3-triazole nitrogen atom, and the side-chain indole NH of Trp181. This interaction may stabilize the bound conformation of the ligand and could account for the observed differences in IC 50 values between the ethers and the acetamides used to link the two halves of the dimer. In addition to the ligand-protein interactions mentioned above, several inter-chain hydrogen bonds are observed between residues in the A and B hGal-3 CRD chains in the complex including Asp239A/ASN179B, Val116A/Asn153B, and Lys233A/Val116B. These additional interactions likely contribute to the stability of the ternary hGal-3 CRD/ dimeric inhibitor complex, and the difficulties obtaining   a In vitro metabolic stability was measured by incubating the ligand in human, rat and mouse liver microsomes and measuring the % remaining after a 10 minute incubation period. HTRF and F-Ligand assays are reported as mean IC 50 values ± SD from at least three experiments unless otherwise noted b Value is the average of two experiments c Value determined from one experiment X-ray crystal structures of hGal-3 CRD with other dimeric inhibitors may be due to reduced ability of those ternary complexes to form inter-chain hydrogen bonds as a result of ligand conformations incompatible with inter-chain proteinprotein contacts. Size-exclusion chromatography combined with multiangle light scattering (SEC-MALS) experiments were utilized to further understand the binding of the monomeric and dimeric Gal-3 inhibitors [37,38]. When observing the hGal-3 protein by itself, the molecular weight (MW) of the protein was 2.7 × 10 4 Da and showed a single peak in the SEC-MALS analysis. Combining the protein with monosaccharide 26, the SEC-MALS analysis showed a single peak with a 2.65 × 10 4 Da MW, consistent with monomeric protein and no change in the oligomerization state. The same experiment was repeated using 10 and the hGal-3 protein, and the protein eluted two minutes earlier from the gel filtration column, consistent with an increase in hydrodynamic radius, and the observed MW increased to 3.6 × 10 4 Da, consistent with a mixture of monomer and dimer. When combining 25 with the hGal-3 protein and performing the SEC-MALS analysis, similar to 10, the peak eluted two minutes earlier and the MW of the major peak increased to 3.6 × 10 4 Da. Both studies involving dimeric ligands showed a small shoulder that stretched back to the initial hGal-3 protein shift and had a molecular weight similar to the unbound protein.
These observations were interpreted to indicate that the monomeric inhibitor 26 was bound to only one hGal-3 protein as expected, while both of the dimeric inhibitors, 10 and 25, showed a different binding profile, one that was likely a mixture of monomeric and dimeric protein interactions based on the peak shift, increase in molecular weight and shoulder stretching back to the original hGal-3 protein peak. While the in vitro assay data, X-ray crystal structure of hGal-3 CRD with dimeric inhibitor 25, and the SEC-MALS results are consistent with the binding of two hGal-3 CRD molecules to dimeric inhibitors in vitro, it is still unknown whether the improved potency achieved with dimeric inhibitors would result in improved in vivo activity.
The majority of Gal-3 antagonists designed to date have targeted monovalent interactions with the galectin protein.
These approaches have mostly focused on glycomimetics aimed at improving the inherent limitations carbohydrates have as drug targets, especially hydrophilicity and lack of metabolic stability [39]. TD139 and GB1211 are evidence that the monovalent glycomimetic approach has been successful at improving the effectiveness of these proteinligand interactions to provide low nanomolar binding affinities while improving ADME properties to deliver druggable targets. In addition to the monovalent Gal-3 antagonist approach, multivalent strategies at Gal-3 antagonism have also been considered and are described in further detail by Bertuzzi et al. [39]. In general, most of these strategies utilize multiple low-affinity binding interactions to enhance the selectivity and overall affinity of these large, complex ligands. The structures presented in this manuscript attempt to bridge the gap between the two strategies and offer dimeric based ligands with exceptional potency and provide evidence of multiple protein-ligand interactions with one ligand.

Conclusions
The X-ray crystal structure analysis of monosaccharide Gal-3 inhibitors bound to the Gal-3 CRD revealed close interligand contacts in the extended crystalline lattices. A set of dimeric compounds was designed to test if improvements in potency compared to the monosaccharide subunits could be achieved using a strategy linking two monosaccharide subunits bound to different Gal-3 CRD molecules. These experiments show that after optimization of linker position, length and atom composition, dimers can improve the potency in HTRF and F-ligand assays when compared to their monosaccharide subunits while maintaining in vitro permeability and metabolic stability. In an optimized case, 10 improved potency 55x in the hGal-3 and 7-13x in the mGal-3 assays respectively when compared to the monosaccharide. Lead compounds 10 and 25 were further examined in in vivo mouse PK studies. PO dosing showed a much lower oral bioavailability compared to the monosaccharide comparators. However, when compound 25 was dosed SQ at 10 mg/kg, AUC total improved from 0.035 μM*h (PO) to 2.23 μM*h. Furthermore, an X-ray crystal structure of 25 bound to hGal3 protein was obtained and crystallography analysis showed the ligand bound to two distinct Gal-3 proteins with each monosaccharide unit occupying the CRD of a single protein. SEC-MALS analysis was consistent with the X-ray crystal structure as the molecular weight and retention time shifted with the introduction of both 10 and 25 with the hGal-3 protein, while combining monosaccharide 26 with the hGal-3 protein did not show a corresponding shift in retention time and molecular weight using the SEC-MALS analysis.

Chemistry
The formation of dimers 3 and 4 proceeded by first forming the C-3 1,2,3-triazole 29 using click chemistry to couple the silyl protected phenyl alkyne 27 and azide 28 (Scheme 1) [40]. The p-methoxybenzylidine acetal group was cleaved to unmask the C-4, C-6-diol 30 by heating in AcOH and H 2 O, then the diol was acylated with acetyl chloride to give 31, where the C-2, C-4, and C-6 alcohols were all protected with as acetates. Treatment of 31 with bromine gave the anomeric bromide 32 in the beta configuration which could then be displaced using sodium azide in DMF at 75°C to give the azide 33 in the anomeric position in the alpha configuration. Global removal of the acetate groups followed using sodium methoxide in MeOH to provide triol 34 which was then subjected to another click reaction, coupling the C-1 azide with alkyl diynes to give linkers 3 and 5 carbons in length between the newly formed triazoles delivering dimers 3 and 4.
Dimer 5 used a different method to link the two halves of the molecule, in this instance extending a linker from the anomeric 1,2,3-triazole to the C-2 alcohol (Scheme 2). Synthesis of the first half of the molecule utilized the C-2 alcohol of 35 [30] as a handle to extend an acetate group using sodium hydride and ethyl bromoacetate, then hydrolyzed to the carboxylic acid 36 by treatment with sodium hydroxide, MeOH, and H 2 O. An amide was formed by treating the carboxylic acid 36 with HATU [41] and an alkynyl amine which extended an alkynyl handle from the C-2 position to give 37. Click chemistry of the alkyne 37 with azide 38 [42] provided 5 which effectively linked the C-2 alcohol of one monosaccharide to the newly formed anomeric 1,2,3-triazole of a second monosaccharide through an acetamide and carbon spacer.
Amide linkages for compounds 7-10 were formed by treating alcohol 39 [30] with sodium hydride followed by t-Bu-bromoacetate to form the t-butyl acetate 40 (Scheme 3). Treatment of 40 with TFA, followed by 1,3-diaminopropane effectively unmasked both the C4, C6-diol and the carboxylic acid groups to give 41. Acetyl chloride was used to protect the C4, C6-diol as the diacetate 42 and the dimers were then formed using HATU and diamines of varied lengths to link the two halves of the molecule through newly formed amide bonds extending from the C-2 position of each half. Removal of the acetate protecting groups with sodium methoxide in MeOH gave the dimers 7-10.
Diamines linked through extended C-2 amides 11-18 and 23-25 were formed using a slightly different synthetic route (Scheme 4). C-2 alcohols 43 [30] and 35, which differ by the substitution pattern on the phenyl group attached to the anomeric 1,2,4-triazole, were treated with sodium hydride followed by ethyl bromoacetate to give the C-2 carboxylates which were hyrolyzed to the C-2 carboxylic acids 36 and 44 using aqueous sodium hydroxide and MeOH. Amide formation was accomplished using HATU and Et 3 N to provide the penultimate intermediates of 11-18 or T3P [43] and Et 3 N for 23-25. Deprotection of the C4, C6-diol was accomplished in each instance by first heating the benzylidine acetals in aqueous AcOH at 70°C, followed by concentration and treatment with K 2 CO 3 in H 2 O and MeOH to give 11-18 and 23-25. Dimers 19 and 20 joined through ether linkages from the C-2 position were formed by treatment of the C-2 alcohols 43 and 45 [30] with sodium hydride, followed by introduction of the diiodoalkane (Scheme 5). Deprotection of the C-4 and C-6 alcohols was accomplished using aqueous AcOH at 70°C followed by concentration and treatment with K 2 CO 3 in H 2 O and MeOH to give dimers 19 and 20.
A one step formation of the carbamate linker to join the two monosaccharide units at the C-2 position was envisioned for 21 and 22, however, in practice the carbamate formation wasn't as cooperative as planned. The pyridyl carbonates 48 and 49 were formed in situ by heating the C-2 alcohols 39 and 43 [30], dipyrin-2-yl carbonate and DMAP in CDCl 3 to 50°C (Scheme 6). The diamino alkanes were introduced directly to the crude reaction mixture, however the process resulted in complex mixture of the dimer, the monomeric amino carbamates 50 and 51 and additional unreacted C-2 alcohol starting materials. To further convert the crude mixtures to the dimer products, a second addition of 48 and 49 were needed. Isolation of carbamate 53 was complicated as purification using reverse phase conditions with an MeCN/ H 2 O /TFA mobile phase cleanly separated the dimer, however, upon concentrating, a portion of the benzylidine acetal group had fallen off. The mixture of protected and deprotected alcohols was treated with aqueous AcOH and heated to 70°C then was treated with K 2 CO 3 in MeOH and H 2 O to give 21. With the lessons learned from the complicated isolation of 53, intermediate 52 was purified using normal phase flash chromatography to avoid the partial deprotection of the benzylidine acetal seen with the reverse phase conditions. Deprotection of the diol was accomplished by heating 52 in AcOH and H 2 O at 70°C followed by concentration and treatment with K 2 CO 3 , MeOH and H 2 O to give 22.

Materials
All reagents were purchased from commercial sources and used without further purification unless otherwise noted. All reactions involving air-or moisture-sensitive reagents were performed under an inert atmosphere. Proton and carbon magnetic resonance ( 1 H and 13 C NMR) spectra were recorded either on a Bruker Avance 400 or a JEOL Eclipse 500 spectrometer and are reported in ppm relative to the reference solvent of the sample in which they were run.  (29) (2 S,4aR,6 S,
(7) 2-(((2 R,3 S,4 R,5 S,6 S)-2-(1-(2,5-dichlorophenyl)-3-methyl-1H-1,2,4-triazol-5-yl)-5-hydroxy-6-(hydroxymethyl)-4-(4-(3,4,5-trifluorophenyl)-1H-1,2,3-triazol-1-yl)tetrahydro-2H-pyran-3-yl)oxy)-N-(2-(2-(((2 S,3 R,4 S,5 R,6 R)-2-(1-(2,5-dichlorophenyl)-3-methyl-1H-1,2,4triazol-5-yl)-5-hydroxy-6-(hydroxymethyl)-4-(4-(3,4,5trifluorophenyl)-1H-1,2,3-triazol-1-yl)tetrahydro-2H-pyran-3-yl)oxy)acetamido)ethyl)acetamide. To a flask containing a solution of 42 (0.036 g, 0.050 mmol) in DMF (2 mL) and triethylamine (0.035 mL, 0.252 mmol) was added HATU (0.058 g, 0.151 mmol) followed by ethylenediamine (1 M in DMF) (0.040 mL, 0.040 mmol). The mixture was stirred at rt for 24 h then was filtered through a plug of glass wool and was purified by preparative HPLC using a C18 column and a H 2 O /MeCN gradient with TFA buffer. The fractions containing the product were concentrated under reduced pressure to give the dimer as a white solid (7 mg, 0.0048 mmol, 19% yield). LCMS: m/e 1452.0 (MH + ), 1.03 min (Method 1). To the dimer (0.007 g, 4.82 µmol) in MeOH (1 mL) was added sodium methoxide (25% solution in MeOH) (1.1 µl, 4.8 µmol). The mixture was stirred at rt for 16 h. Three drops of 1 N HCl was added and the mixture was concentrated under a stream of nitrogen. The mixture was diluted with DMF, filtered through a plug of glass wool and was purified by preparative HPLC using a C18 column and a H 2 O /MeCN gradient with ammonium acetate buffer. Fractions containing the product were combined and concentrated under reduced pressure to give the title product (1 mL) and triethylamine (0.033 mL, 0.238 mmol) was added HATU (0.054 g, 0.143 mmol) followed by 1,3-diaminopropane (1 M in DMF) (0.038 mL, 0.038 mmol). The mixture was stirred at rt for 24 h, then was filtered through a plug of glass wool and was purified by preparative HPLC using a C18 column and a H 2 O /MeCN gradient with TFA buffer. Fractions containing the product were combined and concentrated under reduced pressure to give the dimerized product (7 mg, 0.0047 mmol, 20% yield) as a white solid. LCMS: m/e 1466.1 (MH + ), 1.03 min (Method 1). To a solution of the dimer (0.007 g, 4.7 µmol) in MeOH (1 mL) was added sodium methoxide (25% in MeOH) (1.1 µl, 4.78 µmol) and the mixture was stirred at rt for 16 h. Three drops of 1 N HCl was added and the mixture was concentrated under a stream of nitrogen. The mixture was diluted with DMF, filtered through a plug of glass wool and was purified using a C18 column and a H 2 O /MeCN gradient with ammonium acetate buffer. Fractions containing the product were concentrated under reduced pressure to give the title product (1.9 mg, 0.0015 mmol, 30% yield). LCMS: m/e 1295.86 (MH + ), 1.84 min (Method 2). 1  acetamide. To a flask containing a solution of 42 (0.056 g, 0.078 mmol) in DMF (2 mL) and triethylamine (0.055 mL, 0.392 mmol) was added HATU (0.090 g, 0.235 mmol) followed by 1,5-diaminopentane (9.2 µl, 0.08 mmol). The mixture was stirred at rt for 45 h, then was diluted with H 2 O (15 mL) and extracted with EtOAc (2 × 15 mL). The combined organic layers were washed with H 2 O (3 × 15 mL), then with brine and were dried over magnesium sulfate, filtered and concentrated under reduced pressure. The residue was purified by flash chromatography using a 10-75% EtOAc in hexanes gradient and a 24 g silica gel column. When the product did not elute, the solvent system was changed to a 0-10% MeOH in DCM gradient. The fractions containing the major product were combined and concentrated under reduced pressure to give the dimer (0.02 g, 0.013 mmol, 33% yield) as an off-white film. LCMS: m/e 1494.2 (MH + ), 1.02 min (Method 1). To a solution of the dimer (0.02 g, 0.013 mmol) in MeOH (1.0 mL) was added sodium methoxide (25% solution in MeOH) (3.0 μl, 0.013 mmol) and the mixture was stirred at rt. After 2.5 h of stirring at rt, 5 drops of 1 N HCl were added and the mixture was concentrated under reduced pressure. The residue was diluted with DMF, filtered through a plug of glass wool and was purified by preparative HPLC using a C18 column and a H 2 O /MeCN gradient with ammonium acetate buffer. Fractions containing the products were concentrated under reduced pressure to give the title product A solution of 43b (0.442 g, 0.654 mmol) in THF (10 mL) was cooled to 0°C and sodium hydride (60% dispersion in mineral oil) (0.131 g, 3.27 mmol) was added. The mixture was stirred for 15 min and ethyl bromoacetate (0.291 mL, 2.62 mmol) was added. The mixture was warmed to rt as the ice bath melted and was stirred for 15.5 h. The reaction was carefully quenched with EtOH and the mixture was concentrated under reduced pressure. The residue was diluted with H 2 O (20 mL) and extracted with EtOAc (2 × 30 mL). The organic layers were washed with brine, dried over magnesium sulfate, filtered and concentrated under reduced pressure. The residue was purified by flash chromatography using a 10-80% EtOAc in hexanes gradient and a 40 g silica gel column. Fractions containing the product were combined and concentrated under reduced pressure to give the ethyl ester as an off-white solid (422 mg, 0.554 mmol, 85% . The mixture was stirred at rt for 40 h, then was diluted with 2 mL of H 2 O and was extracted with dichloromethane (3 × 2 mL). The organic layers were dried over sodium sulfate, filtered and concentrated under reduced pressure. The residue was purified by preparative HPLC using a C18 column and a H 2 O /MeCN gradient with TFA buffer. Fractions containing the product were combined and concentrated under reduced pressure to give the dimer (0.023 g, 0.0154 mmol, 56% yield) as a white solid. LCMS: m/e 1489.9 (MH + ), 1.19 min (Method 1). To a solution of the dimer (0.023 g, 0.015 mmol) in AcOH (0.5 mL) was added H 2 O (0.167 mL) and the mixture was heated to 70°C. After heating the mixture for 30 h, it was cooled to rt and was concentrated under a stream of nitrogen. To the residue was added 0.035 g of K 2 CO 3 and the mixture was diluted with 0.5 mL of MeOH and 0.25 mL of H 2 O. The mixture was stirred at rt for 3 days, then was diluted with H 2 O (1 mL) and extracted with dichloromethane (3 × 1 mL). The combined organic layers were dried over sodium sulfate, filtered and concentrated under reduced pressure. The residue was dissolved in DMF, filtered through a plug of glass wool and was purified by preparative HPLC using a C18 column and a H 2 O /MeCN gradient with ammonium acetate buffer. The fractions containing the product were combined and concentrated under reduced pressure to give the title product  .2 mg, 0.164 mmol). The mixture was stirred at rt for 64 h, then was diluted with 2 mL of H 2 O and was extracted with dichloromethane (3 × 2 mL). The organic layers were dried over sodium sulfate, filtered and concentrated under reduced pressure. The residue was purified by flash chromatography using a 0-100% MeOH in dichloromethane gradient and a 24 g silica gel column. Fractions containing the product were combined and concentrated under reduced pressure to give the dimer product (0.022 g, 0.0145, 53% yield) as an off-white solid. LCMS: m/e 1520.1 (MH + ), 1.18 min (Method 1). To a solution of the dimer (22 mg, 0.014 mmol) in AcOH (0.5 mL) was added H 2 O (0.167 mL) and the mixture was heated to 70°C. After 21 h of heating, the mixture was cooled to rt and was concentrated under a stream of nitrogen. To the residue was added 0.035 g of K 2 CO 3 and the mixture was diluted with 0.5 mL of MeOH and 0.25 mL of H 2 O. The mixture was stirred at rt for 20 h, then was diluted with H 2 O (1 mL) and extracted with dichloromethane (3 × 1 mL). The combined organic layers were dried over sodium sulfate, filtered and concentrated under reduced pressure. The residue was dissolved in DMF, filtered through a plug of glass wool and was purified by preparative HPLC using a C18 column and a H 2 O /MeCN gradient with ammonium acetate buffer. Fractions containing the product were combined and concentrated under reduced pressure to give the title product (m, 10H), 6.86 -6.60 (m, 2H), 2.27 (s, 6H), 0.87 (br s, 4H). 100% purity based on two HPLC methods.

In vitro human and mouse Gal-3 HTRF assays
The assays were performed in 384 white Opti plates in three replicates. From the original stocks, 2.5× working stock concentrations of His-tagged recombinant h or mGal-3 and that of B-ASF were prepared. From the working stock, 20 μL of 15 nM His-tagged h or mGal-3 and B-ASF were added to the plates. In negative control wells, one of the binding partners (B-ASF) was replaced with 20 μL of assay buffer. A concentration range of 50× working stocks was prepared for compounds in 100% DMSO. An aliquot of 1 μL compound working stocks was added to each well. The compound was preincubated with 20 μL of h or mGal-3 for 30 min. After adding 20 μL of B-ASF and incubation for additional 1 h, Terbium-labeled anti-His antibody (5 μL/well, 1.0 nM final concentration) was added and incubated for 30 min. Then, streptavidin (2.5 μL/well, 20 nM final concentration) was added and incubated for 1 h. All incubations were conducted at room temperature with gentle shaking at approx. 250 − 300 rpm. The assay plate was read using the homogeneous time-resolved fluorescence screen protocol (excitation wavelength = 340 nm and emission wavelength = 615 nm/665 nm) on an EnVision 2104 multilabel reader. IC 50 values were calculated using Toolset and Curve Master. The F-ligand assay was run in a similar manner using either 384 well or 1536 well plates and fluorescein-conjugated saccharide probe 9 [33] at a final concentration of 79.8 nM prepared from 9.6 μL of the ligand in 4000 μL of buffer.

X-ray crystallography
The protein expression and purification and compound crystallization procedures for the X-ray co-crystal structures have previously been described [32].
PAMPA assay [44] Compounds and controls are utilized as 10 mM stocks in 100% DMSO. Compounds are diluted 1:100 in pH 7.4 or 5.5 donor-well buffer (pION, catalog no. 110151), providing a 100 μM assay solution in 1% DMSO. Compounds diluted in a donorwell buffer were transferred to Whatman Unifilter plates and filtered prior to 200 μL being dispensed into the donor wells of the assay plates (pION, catalog no. 110163). The PAMPA membrane was formed by pipetting 4 μL of the lipid solution (pION, catalog no. 110169) onto the filter plate (VWR CAT #13503). The membrane was then covered with 200 μL of acceptor-well buffer at pH 7.4 (pION, catalog no. 110139). The PAMPA-assay plate (donor and acceptor sides) was combined and allowed to incubate at room temperature for 4 h. The plate was then disassembled, and spectrophotometer plates (VWR, catalog no. 655801) were filled (150 μL/well). The donor, acceptor, reference, and blank plates were read in a SpectraMax UV plate reader. Data were captured by the pION software, which analyzed the spectra and generated Pc values. The control compounds in this assay are as follows (mean ± SD): ketoprofen (pH 5.5, Pc 34.2 ± 3.1 cm −6 /s; pH 7.4, Pc 1.1 ± 0.3 cm−6 / s), metoprolol (pH 5.5, Pc 10.2 ± 1.1 cm−6 /s; pH 7.4, Pc 83.5 ± 4.7 cm−6 /s), and ranitidine (pH 5.5, Pc 0 ± 0 cm−6 /s; pH 7.4, Pc 0.2 ± 0.1 cm−6 /s).
Metabolic stability in liver microsomes [45] The metabolic stability (Metstab) assay evaluates cytochrome P450 (CYP)-mediated metabolism of test compounds in vitro using human and rat microsomes after a 10 min incubation. The incubation was automated on a Biomek FX automation workstation (Beckman Coulter, Fullerton, CA, USA). Each compound was incubated in duplicate in the respective species at a concentration of 0.5 μM. Compounds were received as 3.5 mM solutions in DMSO and were diluted with CH 3 CN to 50 μM before being added to the prewarmed (37°C) microsomal suspension (1 mg/mL) prepared in 100 mM sodium phosphate, pH 7.4, and 6.6 mM MgCl 2 . The reaction was initiated by adding 17 μL of prewarmed 5 mM NADPH in 100 mM sodium phosphate, pH 7.4, into 153 μL of reaction mix. The concentration of DMSO in the incubation mixture was 0.014%. Reaction components were mixed well, and 75 μL was transferred into 150 μL of quench solution at 0 min time point (t0) and again at the 10 min incubation time point (t10). Quenched mixtures were centrifuged at 1500 rpm in an Allegra X-12 centrifuge (Beckman Coulter) for 15 min, and 90 μL of the supernatant was then transferred to a separate 96-well plate for analysis. The metabolism rate was determined based on the parent compound disappearance over time, as measured by LC − MS/MS. Serum protein binding assay [44] The serum protein binding assay was performed using a standard equilibrium dialysis method. In brief, stock solutions of compounds were prepared by solubilizing in 100% DMSO (1 mM). These were spiked in serum so that the final test compound concentration was 10 μM (final DMSO concentration 1%). Serum with the compound was added to one of two chambered rapid equilibrium dialysis (RED) assay plates (8000-dalton molecular weight cut-off), and sodium phosphate buffer was added to the other chamber. Dialysis was performed at 37°C for 5 h in a 10% CO 2 atmosphere. Samples were collected from serum and buffer chambers at pre and postincubation periods. These were analyzed using liquid chromatography with tandem mass spectrometry (LC − MS/MS) to evaluate the fraction of compound (percentage) free to diffuse and equilibrate between the buffer and serum chambers in the RED device. Percent free was calculated using the ratio of compound concentration in buffer to compound concentration in serum at the end of the dialysis period (5 h).

SEC-MALS method
Isocratic separations were performed on a GE Healthcare Superdex 200 Increase 10/300 GL column (10 mm×300 mm), connected to Shimadzu Prominence UFLC in buffer containing HBS, pH 7.4 (Cytiva), with 0.02% Na azide added and 0.1 μm filtered running at a flow rate of 0.75 mL/min. Data were obtained from three online detectors connected in series: A Shimadzu Prominence dual wavelength UV/vis spectrophotometer followed by a Wyatt Technologies DAWN multi-angle laser light scattering detector then a Wyatt Optilab interferometric refractometer. Data were collected and analyzed using ASTRA 8 (Wyatt) and LabSolutions Lite (Shimadzu) software.

Pharmacokinetic studies in mice
The pharmacokinetic studies were conducted at Syngene International Ltd., Biocon-Bristol Myers Squibb Research Center, Bangalore, India, after approval of the Institutional Animal Ethics Committee (IAEC). The facility is registered by the Committee for the Purpose of Control and Supervision on Experiments on Animals (CPCSEA) and accredited to the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). All the animal studies were followed as per the protocol SYNGENE/ IAEC/804/1-2017, approved by the institutional animal ethics committee. The study protocol was previously described [44].