Crystallization of Galectin-8 Linker Reveals Intricate Relationship between the N-terminal Tail and the Linker

Galectin-8 (Gal-8) plays a significant role in normal immunological function as well as in cancer. This lectin contains two carbohydrate recognition domains (CRD) connected by a peptide linker. The N-terminal CRD determines ligand binding specificity, whereas the linker has been proposed to regulate overall Gal-8 function, including multimerization and biological activity. Here, we crystallized the Gal-8 N-terminal CRD with the peptide linker using a crystallization condition that contains Ni2+. The Ni2+ ion was found to be complexed between two CRDs via crystal packing contacts. The coordination between Ni2+ and Asp25 plays an indirect role in determining the structure of β-strand F0 and in influencing the linker conformation which could not be defined due to its dynamic nature. The linker was also shortened in situ and crystallized under a different condition, leading to a higher resolution structure refined to 1.08 Å. This crystal structure allowed definition of a short portion of the linker interacting with the Gal-8 N-terminal tail via ionic interactions and hydrogen bonds. Observation of two Gal-8 N-terminal CRD structures implies that the N-terminal tail and the linker may influence each other’s conformation. In addition, under specific crystallization conditions, glycerol could replace lactose and was observed at the carbohydrate binding site. However, glycerol did not show inhibition activity in hemagglutination assay.


Gal-8_1-186 Induces Erythrocyte Agglutination
The hemagglutination assay shows that truncated Gal-8_1-186 induces agglutination of chicken erythrocytes with a Minimum Agglutination Concentration (MAC) of 0.16 µg/mL ( Figure 1A). Even though this truncated variant of Gal-8 with only one CRD could induce agglutination, previous chemical cross-linking studies did show that it could form dimers [29], and one crystal structure did show that the Gal-8 N CRD did form dimers via crystal packing contacts [22]. This suggests that our truncated Gal-8 may induce erythrocyte agglutination by forming some type of dimeric structure. Furthermore, lactose did inhibit the biological activity of Gal-8_1-186, with a Minimum Inhibition Concentration (MIC) around 0.63 mg/mL for the disaccharide ( Figure 1B). Because one of our structures reported here did bind glycerol (see the following section), we investigated whether glycerol could inhibit Gal-8 N-induced agglutination. Our results demonstrate that glycerol could not inhibit Gal-8 biological activity, indicating, as expected, that glycerol is not a physiological ligand of Gal-8 ( Figure 1C), consistent with a previous report on glycerol and Gal-3 [24].

Gal-8_1-186 Induces Erythrocyte Agglutination
The hemagglutination assay shows that truncated Gal-8_1-186 induces agglutination of chicken erythrocytes with a Minimum Agglutination Concentration (MAC) of 0.16 µg/mL ( Figure 1A). Even though this truncated variant of Gal-8 with only one CRD could induce agglutination, previous chemical cross-linking studies did show that it could form dimers [29], and one crystal structure did show that the Gal-8 N CRD did form dimers via crystal packing contacts [22]. This suggests that our truncated Gal-8 may induce erythrocyte agglutination by forming some type of dimeric structure. Furthermore, lactose did inhibit the biological activity of Gal-8_1-186, with a Minimum Inhibition Concentration (MIC) around 0.63 mg/mL for the disaccharide ( Figure 1B). Because one of our structures reported here did bind glycerol (see the following section), we investigated whether glycerol could inhibit Gal-8 N-induced agglutination. Our results demonstrate that glycerol could not inhibit Gal-8 biological activity, indicating, as expected, that glycerol is not a physiological ligand of Gal-8 ( Figure 1C), consistent with a previous report on glycerol and Gal-3 [24]. 16 µg/mL. "−"represents the negative control well containing no protein; (B) the agglutination capability of Gal-8_1-186 was inhibited by lactose. 0.16 µg/mL Gal-8_1-186 was used in this assay. The MIC (Minimum Inhibition Concentration) of lactose is 0.63 mg/mL. "−"represents the negative control well containing no protein and lactose; "+"represents the positive control well containing no lactose; (C) the agglutination capability of Gal-8_1-186 was inhibited by glycerol. 0.16 µg/mL Gal-8_1-186 was used in this assay. Glycerol cannot inhibit the agglutination activity of Gal-8_1-186 even at 50% (v/v) concentrations. "−" represents the negative control well containing no protein and glycerol; "+" represents the positive control well containing no glycerol.

SDS-PAGE of Protein Crystals
SDS-PAGE of re-dissolved protein crystals from two crystallization conditions demonstrated that the molecular weights of the crystallized proteins are different ( Figure 2). Prior to crystallization trials, the His-tag was removed from the truncated protein by thrombin digestion (i.e., at VPR|GSHM) over 24 h, but the protease was not separated from the Gal-8_1-186 preparation. In this regard, Figure 2A indicates that thrombin may slowly digest Gal-8_1-186, reducing its size. The smaller fragment, which contains the N-terminal CRD and a short piece of the linker, was then crystallized. This shorter variant produced crystals yielding a higher resolution structure (Table 1). The MIC (Minimum Inhibition Concentration) of lactose is 0.63 mg/mL. "−"represents the negative control well containing no protein and lactose; "+"represents the positive control well containing no lactose; (C) the agglutination capability of Gal-8_1-186 was inhibited by glycerol. 0.16 µg/mL Gal-8_1-186 was used in this assay. Glycerol cannot inhibit the agglutination activity of Gal-8_1-186 even at 50% (v/v) concentrations. "−" represents the negative control well containing no protein and glycerol; "+" represents the positive control well containing no glycerol.

SDS-PAGE of Protein Crystals
SDS-PAGE of re-dissolved protein crystals from two crystallization conditions demonstrated that the molecular weights of the crystallized proteins are different ( Figure 2). Prior to crystallization trials, the His-tag was removed from the truncated protein by thrombin digestion (i.e., at VPR|GSHM) over 24 h, but the protease was not separated from the Gal-8_1-186 preparation. In this regard, Figure 2A indicates that thrombin may slowly digest Gal-8_1-186, reducing its size. The smaller fragment, which contains the N-terminal CRD and a short piece of the linker, was then crystallized. This shorter variant produced crystals yielding a higher resolution structure (Table 1).   , 20% (w/v) PEG2000 MME; "2" indicates the protein after his-tag cut off. The protein bands in "1" and "2" are same that means Gal-8 N-terminal CRD with the linker was crystallized.
In Figure 2B, the molecular weight of the crystallized protein was found to be the same before and after crystallization, indicating that the Gal-8 N-terminal CRD with the linker was crystallized. The molecular weight of this protein is approximately 14 kDa; (B) "M" indicates protein molecular marker; "1" indicates the crystallized protein in the second crystallization condition containing 0.1 M Tris pH 8.5, 0.01 M NiCl 2 , 20% (w/v) PEG2000 MME; "2" indicates the protein after his-tag cut off. The protein bands in "1" and "2" are same that means Gal-8 N-terminal CRD with the linker was crystallized.
In Figure 2B, the molecular weight of the crystallized protein was found to be the same before and after crystallization, indicating that the Gal-8 N-terminal CRD with the linker was crystallized. Unfortunately, the structure of the linker could not be defined, because following crystal structure refinement, no electron density for the linker was observed (from Ser154 to end), indicating that the linker is highly flexible.
The best structure, structure 1, allowed 152 residues (Asn7 to Gln158) to be identified with 1.08 Å resolution. This is the best resolution for the Gal-8 N CRD available to date. The initial six residues (Met1 to Asn6) at the N-terminus of the CRD were not visible, likely due to structural disorder within the crystals. Electron density of the first four residues in the linker region (Ser154 to Gln158) was clear, indicating that this part of linker forms stable secondary structure under conditions of our crystallization. In structure 1, four residues (Asn7, Gln9, Asp156, and Gln158) were seen to interact with each other via ionic interactions and hydrogen bonds ( Figure 3A). Although the crystallization conditions were the same for structures 2 and 3, residues from the linker were less visible compared to our 1.08 Å structure. In structures 4 and 5, the electron density of residues from Ser154 to the C-terminus was not visible, suggesting greater flexibility from this segment. In structures 1-3, the N terminal tail (Gln9 to Ile12) forms a rigid loop ( Figure 3B), whereas in structures 4 and 5, the N terminal tail (Gln9 to Ile12) forms a small β-strand F0.
In structures 4 and 5, Ni 2+ formed a complex with Asp25 and His38, His147 from another Gal-8N molecule, along with three water molecules ( Figure 3C). In contrast in structures 1 to 3, no divalent caution was observed at this position. In a previously reported structure, Zn 2+ was found at the same position as Ni 2+ in the crystal structure with space group P4 3 2 1 2 [32]. Asp25 is close to β-strand S1, and coordination between Asp25 and Ni 2+ may impact on S1. Therefore, S1 may induce Gln9 to Ile12 to form a β-strand from Ile12 to Phe19 via van der Waals forces.
It appears that there is a relationship between the N-terminal tail (especially Gln9 to Ile12 region) and the linker. If the N-terminal tail interacts with the linker, then Gln9 to Ile12 forms a rigid loop and part of the linker (here is from Ser154 to Gln158) is stabilized ( Figure 3A). In contrast, if the N-terminal tail cannot interact with the linker for some reason (e.g., due to the Ni 2+ ion mentioned above), then Gln9 to Ile12 forms a β-strand with the linker having high flexibility. This would indicate that the N-terminal tail and the linker could influence the conformation of each other.
Gal-8 variant F19Y is significantly associated with rheumatoid arthritis [13]. If Phe19 were mutated to Tyr, then the hydroxyl group of Tyr could influence the interaction between Phe19 and Ile12, such that the structure of the N-terminal tail would be influenced. After that, the structure of linker should also be changed. Finally, alternation of linker conformation could also result in global effects to the Gal-8 structure and, therefore, function. This prospect is consistent with the previously published results from circular dichroism (CD) [32]. There, a far-UV CD spectrum revealed a slight secondary structure difference between Gal-8 N-terminal CRD and the F19Y variant. [32]. The CD spectrum of thermal denaturation experiments also showed that the T 1/2 of the full-length form of the F19Y variant was 2.5 • C higher than that of the wild type protein. This indicates that the impact from the F19Y substitution on the structure of the lectin is not confined to the N-terminal domain [32]. The N-terminal tail interacts with the bottom of Gal-8 linker (structure 1); (B) Phe19 could interact with Ile12 through van der Waals forces (structure 1: yellow; structure 2: purple; structure 3: pink; structure 4: cyan; structure 5: green); (C) Ni 2+ was found to form a complex with Asp25 and two histidine residues (His38 and His147) from another Gal-8N molecule. Two Gal-8 N-terminal CRDs from two asymmetric units were labeled by green and cyan. This figure was generated by using structure 4 (PDB: 5GZF). One pink ball indicates Ni 2+ . Three green balls indicate water molecules. N-terminal and C-terminal of the structure are close to each other and are labeled by gold. C-terminal of the structure is also the starting point of the linker.

Effect of Cryoprotectants on Gal-8_1-186 Crystals
Surprisingly, crystal structure 1 made using the first condition has space group P212121, and not C2, and the carbohydrate binding site binds glycerol and not lactose (Figure 4). This structure was solved with the crystal being soaked in 20% (v/v) glycerol for approximately 1 min prior to data collection. This indicates that using this high concentration of glycerol changed the space group of crystal from C2 to P212121, while replacing lactose at the carbohydrate binding site. The resolution of this structure is 1.08 Å, which is higher than any other Gal-8 structure solved to date. When using XDS or Aimless, the geometric parameters of our five structures could not be modified using different geometric parameters. yellow; structure 2: purple; structure 3: pink; structure 4: cyan; structure 5: green); (C) Ni 2+ was found to form a complex with Asp25 and two histidine residues (His38 and His147) from another Gal-8N molecule. Two Gal-8 N-terminal CRDs from two asymmetric units were labeled by green and cyan. This figure was generated by using structure 4 (PDB: 5GZF). One pink ball indicates Ni 2+ . Three green balls indicate water molecules. N-terminal and C-terminal of the structure are close to each other and are labeled by gold. C-terminal of the structure is also the starting point of the linker.

Effect of Cryoprotectants on Gal-8_1-186 Crystals
Surprisingly, crystal structure 1 made using the first condition has space group P2 1 2 1 2 1 , and not C2, and the carbohydrate binding site binds glycerol and not lactose (Figure 4). This structure was solved with the crystal being soaked in 20% (v/v) glycerol for approximately 1 min prior to data collection. This indicates that using this high concentration of glycerol changed the space group of crystal from C2 to P2 1 2 1 2 1 , while replacing lactose at the carbohydrate binding site. The resolution of this structure is 1.08 Å, which is higher than any other Gal-8 structure solved to date. When using XDS or Aimless, the geometric parameters of our five structures could not be modified using different geometric parameters. Glycerol can be used as a cryoprotectant in cryoannealing, a technique used to improve the quality of crystal data sets and minimize R values. After several cycles of flash-cooling and room temperature exposure, the quality of crystals presumed to be better [33]. In addition, the geometric parameters of several crystals could be changed by using this technique. Before the Gal-8 N crystal was loaded onto the data collection unit, the crystal was exposed to room temperature for approximately 2 s. Even though this approach is different from cryoannealing, it may accomplish the same result. However, the presence of glycerol did not influence the geometric parameters of the structures from crystals made using the second crystallization condition (Table 1). This indicates that the crystal packing contacts in the P43212 structure were stable.
Glycerol, ethylene glycol, and PEG400 are usually used as cryoprotectants for production of galectin crystals. Here, we compared the effects of cryoprotectants on the Gal-8 N CRD structures. The resolution of C2 structures (structures 2 and 3, cryo-protected by ethylene glycol and PEG400, respectively) are lower than our P212121 structure (structure 1, cryo-protected by glycerol). The Rfree values of the C2 structures are higher than our P212121 structure, especially the Rfree value of the PEG400 structure. This indicates that cryoprotectants do play a role in determining the quality of Gal-8 N crystal structures.

The Carbohydrate Binding Site
In structure 1, glycerol was found to replace lactose in the carbohydrate binding site. The positions of residues Arg45, His65, Asn67, Arg69, Trp86, and Glu89 coordinate with both glycerol and lactose ( Figure 5). However, the side chain of Arg45 in this structure shifts by approximately 0.5 Å compared to that in other structures. In structures of Gal-3, an Arg residue is also located at the same position as Arg 162. This Arg residue has been suggested to coordinate the pyranose ring of lactose in the Gal-3 structure. It seems that this arginine in the Gal-8N CRD can adopt different conformations depending upon which ligand binds to the Gal-8 N CRD.
In our crystal structure, the C and O atoms of glycerol occupy the same positions as C4, C5, C6 and O1, O4, O6 atoms of galactose in the lactose-bound structure ( Figure 5). Wild type Gal-8 N 1-152 has been crystallized in the P43212 space group [13] when glycerol was also used as the cryoprotectant. However, in that study, glycerol did not replace lactose, and was found to be located close to Tyr141 [13]. Moreover, in that instance, glycerol did not influence lactose binding to the Gal-8 N-terminal CRD. In our P43212 structure, glycerol also could not replace lactose. Taken together, this likely indicates that either the binding affinity between lactose and the Gal-8 N CRD is different under different crystallization conditions or that the ability of glycerol to replace lactose in the carbohydrate binding site is different under different crystallization conditions. Under solution conditions used in the hemagglutination assay, Gal-8_1-186 may not have high affinity for glycans on erythrocyte membranes. On the other hand, glycerol could not inhibit erythrocyte agglutination induced by Glycerol can be used as a cryoprotectant in cryoannealing, a technique used to improve the quality of crystal data sets and minimize R values. After several cycles of flash-cooling and room temperature exposure, the quality of crystals presumed to be better [33]. In addition, the geometric parameters of several crystals could be changed by using this technique. Before the Gal-8 N crystal was loaded onto the data collection unit, the crystal was exposed to room temperature for approximately 2 s. Even though this approach is different from cryoannealing, it may accomplish the same result. However, the presence of glycerol did not influence the geometric parameters of the structures from crystals made using the second crystallization condition (Table 1). This indicates that the crystal packing contacts in the P4 3 2 1 2 structure were stable.
Glycerol, ethylene glycol, and PEG400 are usually used as cryoprotectants for production of galectin crystals. Here, we compared the effects of cryoprotectants on the Gal-8 N CRD structures. The resolution of C2 structures (structures 2 and 3, cryo-protected by ethylene glycol and PEG400, respectively) are lower than our P2 1 2 1 2 1 structure (structure 1, cryo-protected by glycerol). The R free values of the C2 structures are higher than our P2 1 2 1 2 1 structure, especially the R free value of the PEG400 structure. This indicates that cryoprotectants do play a role in determining the quality of Gal-8 N crystal structures.

The Carbohydrate Binding Site
In structure 1, glycerol was found to replace lactose in the carbohydrate binding site. The positions of residues Arg45, His65, Asn67, Arg69, Trp86, and Glu89 coordinate with both glycerol and lactose ( Figure 5). However, the side chain of Arg45 in this structure shifts by approximately 0.5 Å compared to that in other structures. In structures of Gal-3, an Arg residue is also located at the same position as Arg 162. This Arg residue has been suggested to coordinate the pyranose ring of lactose in the Gal-3 structure. It seems that this arginine in the Gal-8N CRD can adopt different conformations depending upon which ligand binds to the Gal-8 N CRD.
In our crystal structure, the C and O atoms of glycerol occupy the same positions as C4, C5, C6 and O1, O4, O6 atoms of galactose in the lactose-bound structure ( Figure 5). Wild type Gal-8 N 1-152 has been crystallized in the P4 3 2 1 2 space group [13] when glycerol was also used as the cryoprotectant. However, in that study, glycerol did not replace lactose, and was found to be located close to Tyr141 [13]. Moreover, in that instance, glycerol did not influence lactose binding to the Gal-8 N-terminal CRD. In our P4 3 2 1 2 structure, glycerol also could not replace lactose. Taken together, this likely indicates that either the binding affinity between lactose and the Gal-8 N CRD is different under different crystallization conditions or that the ability of glycerol to replace lactose in the carbohydrate binding site is different under different crystallization conditions. Under solution conditions used in the hemagglutination assay, Gal-8_1-186 may not have high affinity for glycans on erythrocyte membranes. On the other hand, glycerol could not inhibit erythrocyte agglutination induced by Gal-8_1-186. Furthermore, the Gal-8 N-terminal CRD in solid (crystallized) and liquid states may have different affinities towards various ligands. Therefore, glycerol could replace lactose in the Gal-8 N-terminal CRD in the crystallized form, and could not replace competition with carbohydrates in the agglutination assay. Gal-8_1-186. Furthermore, the Gal-8 N-terminal CRD in solid (crystallized) and liquid states may have different affinities towards various ligands. Therefore, glycerol could replace lactose in the Gal-8 N-terminal CRD in the crystallized form, and could not replace competition with carbohydrates in the agglutination assay. Collins et al. reported that the time scale of glycerol displacement for lactose in Gal-3 crystals is on the order of days [23]. Glycerol is a weak ligand for Gal-3 [24], such that high affinity ligands should readily displace glycerol. This approach has been suggested as a good method to exchange ligands with Gal-3. Here, we observed that the lactose displacement time with glycerol in Gal-8 N-terminal CRD crystals was only ~1 min, much shorter than that found with Gal-3. This suggests that with Gal-8 N, it is feasible to use this approach for ligand exchange.

Discussion
Human tandem-repeat-type lectins (Gal-4, -8, -9, and -12) contain two CRDs, whose structures are highly homologous. However, the linker connecting their N-and C-CRDs, as well as the short tail preceding the N-CRD, is highly variable. This variance likely determines a given lectins' specific functional role. For example, these linkers play roles in interactions with other proteins, membrane insertions, and the regulation of CRD presentations. On the other hand, these diverse linkers are also involved in strengthening signaling potency by intermolecular interactions between CRDs and through multimerization that increases lattice formation on the cell surface [34]. However, the conformation of these linkers remains unknown due likely to their high flexibility. Here, we have successfully crystallized the Gal-8 N-terminal CRD with its linker under crystallization conditions that contain 0.01 M Ni 2+ , a condition that is known to induce trimerization of Gal-9 CRDs [35]. Although the structure of the complete linker was not determined in our structures, the information that we did uncover may provide clues to explain the functions of Gal-8.
In this study, we did not separate thrombin from Gal-8_1-186, primarily because the protease may play a role in Gal-8 L function. In a previous report with Gal-8L, thrombin could digest Gal-8 L, but not Gal-8 M [36]. In that study, however, the digestion time was shorter than the crystallization time (at least two weeks) used here, and the digestion site was assumed to be Arg-Gly and Arg-Thr in the linker region of Gal-8 L. Here, we based the Gal-8 truncation of the linker on Gal-8 M, where the linker region does not contain the RG and RT recognition element. However, thrombin may still have digested our truncated protein. Actually, thrombin is not a digestion site-specific protease, so it Collins et al. reported that the time scale of glycerol displacement for lactose in Gal-3 crystals is on the order of days [23]. Glycerol is a weak ligand for Gal-3 [24], such that high affinity ligands should readily displace glycerol. This approach has been suggested as a good method to exchange ligands with Gal-3. Here, we observed that the lactose displacement time with glycerol in Gal-8 N-terminal CRD crystals was only~1 min, much shorter than that found with Gal-3. This suggests that with Gal-8 N, it is feasible to use this approach for ligand exchange.

Discussion
Human tandem-repeat-type lectins (Gal-4, -8, -9, and -12) contain two CRDs, whose structures are highly homologous. However, the linker connecting their N-and C-CRDs, as well as the short tail preceding the N-CRD, is highly variable. This variance likely determines a given lectins' specific functional role. For example, these linkers play roles in interactions with other proteins, membrane insertions, and the regulation of CRD presentations. On the other hand, these diverse linkers are also involved in strengthening signaling potency by intermolecular interactions between CRDs and through multimerization that increases lattice formation on the cell surface [34]. However, the conformation of these linkers remains unknown due likely to their high flexibility. Here, we have successfully crystallized the Gal-8 N-terminal CRD with its linker under crystallization conditions that contain 0.01 M Ni 2+ , a condition that is known to induce trimerization of Gal-9 CRDs [35]. Although the structure of the complete linker was not determined in our structures, the information that we did uncover may provide clues to explain the functions of Gal-8.
In this study, we did not separate thrombin from Gal-8_1-186, primarily because the protease may play a role in Gal-8 L function. In a previous report with Gal-8L, thrombin could digest Gal-8 L, but not Gal-8 M [36]. In that study, however, the digestion time was shorter than the crystallization time (at least two weeks) used here, and the digestion site was assumed to be Arg-Gly and Arg-Thr in the linker region of Gal-8 L. Here, we based the Gal-8 truncation of the linker on Gal-8 M, where the linker region does not contain the RG and RT recognition element. However, thrombin may still have digested our truncated protein. Actually, thrombin is not a digestion site-specific protease, so it could digest some sites non-specifically [37]. Long incubation times with thrombin could explain why only part of the linker was identified using our condition 1 (i.e., 20% (w/v) PEG2000 MME). Nevertheless, we could observe that this part of the linker did interact with the N-terminus of Gal-8 N via hydrogen bond formation. This is consistent with the predicted model of Gal-8, which suggests that the linker can sometimes form a secondary structure [38].
Frontal affinity gel chromatography [20] and fluorescence polarization have shown that the ligand binding specificities of the two Gal-8 CRDs are different. The N-terminal CRD prefers sulfated and sialylated glycans, whereas the C-terminal CRD prefers blood group A/B [39]. In particular, the N-terminal CRD strongly binds to sulfated and sialylated glycans. The crystal structure of the N-terminal CRD bound with sulfated and sialylated glycans has been solved [22], and it has been suggested that the N-terminal CRD determines the function of Gal-8 [21]. In this present report, one of our structures shows that glycerol binds to the carbohydrate binding site. Therefore, it may be a good idea to substitute lactose in Gal-8 N CRD crystals with glycerol and then replace glycerol with a high affinity ligand.

Crystallization, Data Collection, and Structure Determination
Hampton Research (Aliso Viejo, CA, USA) packs (PEGRx1, PEGRx2, Index, Crystal Screen, and Crystal Screen 2) were used for the initial crystallization screen (sitting-drop vapor diffusion method). After four weeks of incubation at 25 • C, small crystals had formed under the conditions of Crystal screen 2 No. 45 (0.1 M Tris pH 8.5, 0.01 M NiCl 2 , 20% (w/v) PEG2000 MME). To obtain a crystal that was suitable for X-ray diffraction, we used the hanging-drop and sitting-drop methods. Larger crystals were obtained after about two weeks from drops that contained 1 µL protein (12 mg/mL) and 1 µL solution from the well containing 20% (w/v) PEG2000 MME (hanging-drop method) or 0.1 M Tris pH 8.5, 0.01 M NiCl 2 , 20% (w/v) PEG2000 MME (sitting-drop method) at 25 • C.
The crystals were fished out from the crystallization drops and were washed by the reservoir solution. The crystallized proteins were identified by SDS-PAGE with coomassie blue staining.
Prior to X-ray data collection, crystals were soaked for approximately 1 min in the reservoir solution supplemented with 20% (v/v) glycerol, etheglycol or PEG400 as a cryoprotectant, and then flash cooled in liquid nitrogen. All crystals were grown at the same temperature. Data sets were collected at 100 K at the Shanghai Synchrotron Radiation Facility (Shanghai, China).
Data sets were indexed and integrated using XDS (MPI for Medical Research, Heidelberg, Germany) [42] and scaled using Aimless [43] from the CCP4 package (6.4.0, Science and Technology Facilities Council, Rutherford Appleton Laboratory, Didcot, Oxon, UK, 2015) [44]. Structures were determined by Phaser (University of Cambridge, Cambridge, UK) [45] with a molecular replacement method using the structure of Gal-8 variant (PDB: 4BME) as the search model. Structure refinement and water updating were performed using Phenix (Lawrence Berkeley Laboratory, Berkeley, CA, USA) [46] refine and manual adjustment. Final structure validations were performed using MolProbity (Duke University, Durham, NC, USA) [47,48]. Figures for all structures were generated using PyMOL (Schrödinger K.K, Cambridge, MA, USA) [49] or Coot (MRC Laboratory of Molecular Biology, Cambridge, UK) [50].

Hemagglutination Assay
The hemagglutination assay was performed according to Gao et al. [51]. With some minor modifications. Chicken erythrocytes were prepared from fresh blood collected in Alsever's medium (2.05% glucose, 0.8% sodium citrate, 0.42% sodium chloride, and 0.055% citric acid) and were washed four times with 0.15 M NaCl. Cells were then suspended in 4% (v/v) PBS (pH 7.4) containing 1 mg/mL trypsin and incubated at 37 • C for 1 h. After washing with 0.15 M NaCl, the cells were fixed for 1 h in PBS (pH 7.4) containing 1% glutaraldehyde at room temperature followed by termination with five volumes of 0.1 M glycine in PBS (pH 7.4). The fixed cells were washed and adjusted to 10% (v/v) with PBS (pH 7.4). Cells were maintained at 4 • C until use. The hemagglutination assay was performed in microtiter V plates, with each well containing Gal-8_1-186 in 75 µL Tris buffer (10 mM Tris-HCl, 150 mM NaCl, pH 8.0) or 75 µL Tris buffer and 25 µL 4% (v/v) chicken erythrocyte suspensions. The cells were added last, followed by shaking, and agglutination was allowed to proceed for 60 min on the ice to ensure the temperature was consistent. Lactose and glycerol were used to inhibit the biological activity of Gal-8_1-186 protein.
Author Contributions: Yunlong Si and Yue Wang participated in protein overexpression and solving Gal-8 structures; Yunlong Si ran the SDS-PAGE; Yue Wang, Chenyang Song, Jin Gao, Runjie Wu, and Shiqiong Feng participated in Hemagglutination Assay; Jiyong Su, Guihua Tai, and Yifa Zhou conceived of the study, participated in its design and coordination, and helped to draft the manuscript; All authors read and approved the final manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.