Galectin-8 binds to the Farnesylated C-terminus of K-Ras4B and Modifies Ras/ERK Signaling and Migration in Pancreatic and Lung Carcinoma Cells

K-Ras is the most prominent driver of oncogenesis and no effective K-Ras inhibitors have been established despite decades of intensive research. Identifying new K-Ras-binding proteins and their interaction domains offers the opportunity for defining new approaches in tackling oncogenic K-Ras. We have identified Galectin-8 as a novel, direct binding protein for K-Ras4B by mass spectrometry analyses and protein interaction studies. Galectin-8 is a tandem-repeat Galectin and it is widely expressed in lung and pancreatic carcinoma cells. siRNA-mediated depletion of Galectin-8 resulted in increased K-Ras4B content and ERK1/2 activity in lung and pancreatic carcinoma cells. Moreover, cell migration and cell proliferation were inhibited by the depletion of Galectin-8. The K-Ras4B–Galectin-8 interaction is indispensably associated with the farnesylation of K-Ras4B. The lysine-rich polybasic domain (PBD), a region that is unique for K-Ras4B as compared to H- and N-Ras, stabilizes the interaction and accounts for the specificity. Binding assays with the deletion mutants of Galectin-8, comprising either of the two carbohydrate recognition domains (CRD), revealed that K-Ras4B only interacts with the N-CRD, but not with the C-CRD. Structural modeling uncovers a potential binding pocket for the hydrophobic farnesyl chain of K-Ras4B and a cluster of negatively charged amino acids for interaction with the positively charged lysine residues in the N-CRD. Our results demonstrate that Galectin-8 is a new binding partner for K-Ras4B and it interacts via the N-CRD with the farnesylated PBD of K-Ras, thereby modulating the K-Ras effector pathways as well as cell proliferation and migration.

K-Ras (G12V), H-Ras (G12V), and N-Ras (G12V) were transiently expressed as EGFP fusion proteins in HEK293 cells, precipitated by anti-GFP µMACS beads, and incubated with PANC-1 cell lysates to precipitate endogenous Galectin-8 long and short from these lysates to investigate whether Galectin-8 also interacted differentially with EGFP-Ras isoforms. The western blot in Figure 1C demonstrates that Galectin-8 isoforms were both markedly precipitated with EGFP-K-Ras (G12V) and to a lower extent with EGFP-H-Ras (G12V) (lanes 1 and 2). Galectin-8 hardly ever interacted with EGFP-N-Ras (G12V) (lane 3) and it did not interact with EGFP (lane 4). Co-immunoprecipitation analyses in which both the EGFP-Ras isoforms and Galectin-8 long or short were co-expressed in HEK293 cells confirmed these results were confirmed and the anti-GFP antibody precipitated EGFP-Ras proteins (Supplementary Figure S2A). A quantitative analysis of the interaction of Galectin-8 from PANC-1 cells with ectopically expressed EGFP-Ras isoforms is presented in Supplementary Figure S2B. EGFP-H-Ras (G12V) showed an interaction of 33.78 ± 15.14% and EGFP-N-Ras (G12V) of 3.44 ± 5.12% (densitometric analyses of n = 4) when setting the interaction of Galectin-8 short with EGFP-K-Ras (G12V) to 100% as reference. Note that we could detect a very weak binding of Galectin-3 (black bar) and an apparent binding of Galectin-1 (grey bar) to EGFP-K-Ras (G12V) in the same precipitate.
Co-immunoprecipitation assays were performed to detect K-Ras-Galectin-8 interaction in living cells since all these analyses were carried out with ectopically expressed Ras proteins. PANC-1 cells, which express considerable amounts of both Galectin-8 isoforms as well as oncogenic and wildtype K-Ras4B [34] were used. Endogenous K-Ras was precipitated while using a K-Ras4B specific antibody and endogenous Galectin-8 was clearly co-precipitated ( Figure 1D). Moreover, cell fractionation assays (see Figure 2E) and immunofluorescence studies (see Supplementary Figure S4A) revealed that Galectin-8 located to the cytoplasm and markedly to the membrane, where it co-localized with K-Ras. Subsequently, we investigated whether purified K-Ras and purified Galectin-8 directly interact with each other in immunodot blot experiments. Therefore, purified His-tagged K-Ras (G12V) and His-tagged Galectin-8 long proteins, expressed as modified proteins in baculovirus-infected insect cells, and unmodified His-tagged K-Ras as well as Galectin-8 short (rec.Gal-8s) from E. coli were used as bait or prey, respectively. Figure 1E demonstrates direct interaction of these proteins. The upper blot in each panel shows the detection of the bound and thus interacting prey protein, and the lower blot the amount of the spotted bait protein. Recombinant Gal-8s interacted with post-translationally modified K-Ras expressed in insect cells (upper panel left). Interestingly, it did not with unmodified K-Ras from E. coli (right panel), where rec.Gal-8s was spotted in increasing amounts as a control. In the inverse experiment, substantial amounts of K-Ras bound to insect-expressed Gal-8l when Gal-8l was spotted and incubated with posttranslationally modified K-Ras (G12V) (lower panel left). Taken together, these results demonstrate that both isoforms of Galectin-8 represent new direct binding proteins for oncogenic K-Ras.

Downregulation of Galectin-8 Affects K-Ras and ERK1/2
The effects of Galectin-8 silencing on the activity of the mitogen-activated protein kinases ERK1/2 were analyzed following the siRNA approach while using a pool of four siRNAs targeting different sequences of Galectin-8 to characterize signal transduction mechanisms regulated by Galectin-8. Figure 2A shows that the application of these siRNAs resulted in a >70% depletion of both Galectin-8 isoforms in PANC-1, EGFP-K-Ras (G12V), and EGFP-expressing PANC-1 cells. PANC-1 cells stably expressing EGFP-K-Ras (G12V) (clone 4.1 and 4.4) (34) were used to detect the effects of Galectin-8 depletion, particularly on EGFP-K-Ras. The silencing of Galectin-8 leading to a two-fold upregulation of the phosphorylation of ERK1/2 in all analyzed cell clones, without affecting the total amount of ERK1/2, is of note ( Figure 2B). In the case of EGFP-K-Ras (G12V) cells, a significant increase in the amount of EGFP-K-Ras (2.87 ± 0.72-fold) as compared to the mock-transfected control accompanied this upregulation. The amount of EGFP did not change (1.12 ± 0.05-fold) ( Figure 2C). The deletion of Galectin-8 in Colo699 lung carcinoma cells, which also express Galectin-8, -3, and -1 as PANC-1 cells (Supplementary Figure S1B), produced a similar increase in ERK1/2 phosphorylation, as shown in Figure 2D. A549 lung carcinoma cells also exhibited a 1.7-fold increased ERK phosphorylation after Galectin-8 depletion, demonstrating a cell line independent effect. The analysis of the subcellular localization of Galectin-8 in PANC-1 cell clones revealed that a substantial amount is localized in the P100 particulate, membrane-containing fraction. EGFP-K-Ras (G12V) is particularly localized in the P100 fraction, whereas EGFP is only restricted to the cytosolic fraction ( Figure 2E, upper panel). Moreover, siRNA-mediated deletion of Galectin-8 caused an increase of EGFP-K-Ras (G12V), especially in the P100 fraction ( Figure 2E, lanes 5 and 6), which represents K-Ras molecules that are capable of inducing signal transduction leading to ERK phosphorylation. The determination of the activity of EGFP-K-Ras (G12V) after Galectin-8 silencing by GST-RBD pull-down assays confirmed a marked increase in the amount of the GTP-bound and, thus, active protein concomitant with an increase in the total amount of EGFP-K-Ras (G12V).
Cancers 2019, 11, x 6 of 24 exhibited a 1.7-fold increased ERK phosphorylation after Galectin-8 depletion, demonstrating a cell line independent effect. The analysis of the subcellular localization of Galectin-8 in PANC-1 cell clones revealed that a substantial amount is localized in the P100 particulate, membrane-containing fraction. EGFP-K-Ras (G12V) is particularly localized in the P100 fraction, whereas EGFP is only restricted to the cytosolic fraction ( Figure 2E, upper panel). Moreover, siRNA-mediated deletion of Galectin-8 caused an increase of EGFP-K-Ras (G12V), especially in the P100 fraction ( Figure 2E, lanes 5 and 6), which represents K-Ras molecules that are capable of inducing signal transduction leading to ERK phosphorylation. The determination of the activity of EGFP-K-Ras (G12V) after Galectin-8 silencing by GST-RBD pull-down assays confirmed a marked increase in the amount of the GTP-bound and, thus, active protein concomitant with an increase in the total amount of EGFP-K-Ras (G12V).  System. The bar graph shows the densitometric analysis of the ratio of pERK1/2 to ERK1/2 normalized to the control (mean ± SEM; * p ≤ 0.05; ** p ≤ 0.01, n = 6). (C) Expression of EGFP-K-Ras and EGFP was evaluated using anti-GFP antibody. GAPDH (lower panel) was determined as a loading control. The diagram shows the densitometric analysis as a ratio of EGFP/GAPDH normalized to the mock-transfected control (mean ± SEM; * p ≤ 0.05, n = 6). (D) Depletion of Galectin-8 in lung carcinoma cells. Colo699 cells were treated as in (A). The western blots show the amount of Galectin-8 and GAPDH (upper panel) and pERK1/2 and ERK1/2 (lower panel). The diagram shows the densitometric analysis of the ratio of pERK1/2 to ERK1/2 (mean ± SEM; * p ≤ 0.05, n = 4). (E) Subcellular localization of EGFP-K-Ras (G12V) and Galectin-8. To determine whether enhanced ERK phosphorylation is associated with an enhanced amount of active, membrane-associated EGFP-K-Ras (G12V), EGFP-14 and K-Ras-4.1 PANC-1 cells were treated as in (A). Lysates were fractionated into soluble [S100] (cytosolic) and particulate [P100] (membranous) fractions and they were analyzed by western blot (S100: 50 µg, P100: 25 µg). The upper blot shows the distribution and depletion of Galectin-8 in both fractions, the middle panel displays the distribution of EGFP-K-Ras, especially in the P100 fraction, and the lower panel shows the enrichment of EGFP in the S100 fraction (n = 4). (membranous) fractions and they were analyzed by western blot (S100: 50 µg, P100: 25 µg). The upper blot shows the distribution and depletion of Galectin-8 in both fractions, the middle panel displays the distribution of EGFP-K-Ras, especially in the P100 fraction, and the lower panel shows the enrichment of EGFP in the S100 fraction (n = 4). The increase in the amount of EGFP-K-Ras (G12V) and in the activity of ERK1/2 was dependent on the Galectin-8 siRNA concentration, as shown in Figure 3A. Based on these observations, we hypothesized that Galectin-8 might function as a novel regulator for maintaining a certain steady-state level of K-Ras within the cell. The mode of K-Ras degradation is controversial. Lu et al. reported that K-Ras undergoes lysosomal degradation [35], but Koo et al. described proteasome-mediated degradation [36]. K-Ras exhibits a half-life of >12 h under normal physiological conditions [37]. Mock and siGal-8 treated PANC-1/EGFP-K-Ras (G12V) cells were treated for 24 h or 48 h with the proteasome inhibitor MG132 to address the issue if inhibition of the proteasome with MG132 affects EGFP-K-Ras or Galectin-8 protein stability. As shown in Figure 3B, MG132 treatment increases the amount of EGFP-K-Ras (white columns) in a time-dependent manner as compared to control. Within 24 h of treatment, this effect was equal or slightly higher than treatment with siGal-8 alone (black bar [control] as compared to white bar [24 h MG132]). The treatment with MG132 for 48 h resulted in a marked and similar increase in EGFP-K-Ras in both mock and siGal-8-treated cells. MG132 did not affect the amount of Galectin-8, whereas the level of β-catenin clearly increased, as expected from numerous studies in the literature. The increase in the amount of EGFP-K-Ras (G12V) and in the activity of ERK1/2 was dependent on the Galectin-8 siRNA concentration, as shown in Figure 3A. Based on these observations, we hypothesized that Galectin-8 might function as a novel regulator for maintaining a certain steady-state level of K-Ras within the cell. The mode of K-Ras degradation is controversial. Lu et al. reported that K-Ras undergoes lysosomal degradation [35], but Koo et al. described proteasome-mediated degradation [36]. K-Ras exhibits a half-life of >12 h under normal physiological conditions [37]. Mock and siGal-8 treated PANC-1/EGFP-K-Ras (G12V) cells were treated for 24 h or 48 h with the proteasome inhibitor MG132 to address the issue if inhibition of the proteasome with MG132 affects EGFP-K-Ras or Galectin-8 protein stability. As shown in Figure 3B, MG132 treatment increases the amount of EGFP-K-Ras (white columns) in a time-dependent manner as compared to control. Within 24 h of treatment, this effect was equal or slightly higher than treatment with siGal-8 alone (black bar [control] as compared to white bar [24 h MG132]). The treatment with MG132 for 48 h resulted in a marked and similar increase in EGFP-K-Ras in both mock and siGal-8-treated cells. MG132 did not affect the amount of Galectin-8, whereas the level of β-catenin clearly increased, as expected from numerous studies in the literature.  Detection of β-catenin served as a control for the activity of MG132. GAPDH served as a loading control. The bar graph shows the mean ± SEM of the densitometric analysis of EGFP-K-Ras normalized to the mock-transfected control without MG132 (n = 3).

Impact of Galectin-8 Deletion on Cell Migration and Proliferation
Wound healing and proliferation experiments were performed after the silencing of Galectin-8 expression to elucidate whether Galectin-8 contributes to the development of a malignant phenotype by affecting cell migration or cell proliferation. Figure 4A demonstrates that down-regulation of Galectin-8 significantly inhibited the migration of all PANC-1 cell clones. Normalization of the velocity to the mock transfected control cells (set to 100%) disclosed a migration rate of 79.3 ± 2.73% for siGal-8-treated EGFP-K-Ras-4.1, 67.3 ± 4.68% for EGFP-K-Ras-4.4, 69.88 ± 5.42% for PANC-1 cells, and 72.05 ± 8.26% for EGFP-14. The depletion of Galectin-8 in A549 lung carcinoma cells resulted in a similar reduction with 71.06 ± 5.27% as compared to the control. Thus, an approximately 30% inhibition of cell migration was discovered in pancreatic and lung carcinoma cells. Moreover, Galectin-8 depletion also reduced cell proliferation.  Figure 4B, bar graph). Again, Galectin-8 was markedly downregulated, as demonstrated by the western blot in Figure 4B. Wound healing and proliferation experiments were performed after the silencing of Galectin-8 expression to elucidate whether Galectin-8 contributes to the development of a malignant phenotype by affecting cell migration or cell proliferation. Figure 4A demonstrates that down-regulation of Galectin-8 significantly inhibited the migration of all PANC-1 cell clones. Normalization of the velocity to the mock transfected control cells (set to 100%) disclosed a migration rate of 79.3 ± 2.73% for siGal-8-treated EGFP-K-Ras-4.1, 67.3 ± 4.68% for EGFP-K-Ras-4.4, 69.88 ± 5.42% for PANC-1 cells, and 72.05 ± 8.26% for EGFP-14. The depletion of Galectin-8 in A549 lung carcinoma cells resulted in a similar reduction with 71.06 ± 5.27% as compared to the control. Thus, an approximately 30% inhibition of cell migration was discovered in pancreatic and lung carcinoma cells. Moreover, Galectin-8 depletion also reduced cell proliferation.  Figure 4B, bar graph). Again, Galectin-8 was markedly downregulated, as demonstrated by the western blot in Figure 4B.

The N-CRD of Galectin-8 Interacts with Farnesylated K-Ras
Galectin-8 belongs to the tandem-repeat Galectins with two CRDs. The N-and C-terminal CRDs of Galectin-8 share 35% homology and might both serve as the binding site for K-Ras. Deletion mutants of Galectin-8, comprising each CRD either alone or in conjunction with the hinge region of Galectin-8 long, were expressed as HA-tagged proteins in HEK293 cells. Figure 5A shows the results of binding experiments with EGFP-K-Ras (G12V). Only the N-CRD, but not the C-CRD was co-precipitated with EGFP-K-Ras (G12V). The adding of the 42 aa hinge region did not alter the interaction of the N-CRD with K-Ras, but it slightly increased binding of the C-CRD-hinge (interaction index: 0.188 ± 0.065) ( Figure 5A). Densitometric analysis determined the interaction index, relating the amount of precipitated HA-CRD to the amount of precipitated EGFP-K-Ras (G12V) and normalized to HA-N-CRD. Next, we analyzed whether the observed specificity for the Ras isoforms was also evident when using N-CRD/N-CRD hinge proteins. As shown in Figure 5B, N-CRD markedly co-precipitated with EGFP-K-Ras (G12V), less with EGFP-H-Ras (G12V), and no interaction was observed with EGFP-N-Ras (G12V). Therefore, the N-CRD mediates the preferential binding of Galectin-8 to K-Ras. Cellular fractionation experiments after the ectopic expression of HA-N-CRD and HA-C-CRD in PANC-1 cells demonstrated that only the N-CRD was markedly detectable in the particulate, membrane-containing fraction co-localizing with endogenous Ras Immunofluorescence analysis of the localization of HA-N-CRD and HA-C-CRD ectopically expressed in HEK293 cells confirmed these results, as shown in Supplementary Figure S4B.
The bound nucleotide in Ras changes the confirmation of the protein and its properties to interact with effector and regulatory proteins [23,38]. This leads to the question of how the activity of K-Ras influences its binding to Galectin-8. Constitutive active GTP-bound EGFP-K-Ras (G12V), dominant negative GDP-bound EGFP-K-Ras (S17N), and wildtype EGFP-K-Ras as the control, each being expressed in HEK293 cells, were incubated with PANC-1 cell lysate for precipitation of Galectin-8. Ras activity pull-down assays verified the activity of the used K-Ras mutants (data not shown). The co-precipitation experiments ( Figure 5C) revealed no difference in the precipitation of Galectin-8; therefore, the bound nucleotide and the activity of K-Ras do not influence the binding of Galectin-8. These results point towards an interaction domain outside of the switch I and switch II region of K-Ras. We predicted that the farnesylation of K-Ras might be crucial for the interaction in view of the fact that we could not detect any binding of Galectin-8 to unmodified K-Ras expressed in E. coli in the previous experiments (see Figure 1E). Precipitation experiments with EGFP-K-Ras (G12V) and EGFP-K-Ras (G12V,C185S) expressed in HEK293 cells, as well as unmodified EGFP-K-Ras (G12V) expressed in E. coli, were performed, to prove this hypothesis. The mutation of the C-terminal cysteine to serine (C185S) in K-Ras abrogates K-Ras farnesylation, which leads to cytosolic localization. Hancock and coworkers first described this [8] and demonstrated in our group by its localization in the soluble fraction in cellular fractionation experiments. Figure 5D-F illustrate that Galectin-8 from PANC-1 cell lysates and ectopically expressed N-CRD interacted exclusively with the farnesylated EGFP-K-Ras and not with the farnesyl-deficient mutant or the recombinant K-Ras. None of the K-Ras proteins interacted with the C-CRD fragment. These data clearly demonstrates that the farnesyl moiety of K-Ras is crucial for the interaction with Galectin-8 via its N-CRD.

The Two Lysine Residues in the Farnesylated Domain of K-Ras Promote the Interaction with Galectin-8
The C-terminal domain of K-Ras is unique with regard to its hexa-lysine polybasic domain adjacent to the farnesylated domain with the characteristic cysteine of the CAAX motif (see Figure 6). Moreover, K-Ras is unique, as it is not palmitoylated within the hypervariable region. These characteristics prompted us to investigate the importance of the lysine residues within the farnesylated domain. Lysine K182 and K184 in K-Ras were mutated to the corresponding serine in H-Ras and the corresponding proline in N-Ras (also see Figure 6 for an overview). Co-precipitation assays, as illustrated in Figure 7A-C, showed that only exchanging one of these lysines did not significantly alter the binding of Galectin-8. However, replacing both lysines significantly reduced the interaction index of Galectin-8 long for K-Ras to 0.49 ± 0.37 (p = 0.038) and of Galectin-8 short to 0.56 ± 0.35 (p = 0.047) as compared to the wildtype K-Ras form ( Figure 7A). Introducing a lysine in position 183 of H-Ras, resulting in two lysine residues next to the farnesylated cysteine, increased the interaction index of both Galectin-8 isoforms for H-Ras by <1.5-fold, as compared to the wildtype form of H-Ras ( Figure 7B). However, exchanging K185 in H-Ras to proline did not alter the binding index, which indicated that this lysine alone is not critical for the interaction. N-Ras exhibit no interaction with Galectin-8, but introducing lysine residues in position 183 and 185 resulted in a detectable amount of co-precipitated Galectin-8 short. The insertion of K185 alone produced a weaker interaction with Galectin-8 short.

The Two Lysine Residues in the Farnesylated Domain of K-Ras Promote the Interaction with Galectin-8
The C-terminal domain of K-Ras is unique with regard to its hexa-lysine polybasic domain adjacent to the farnesylated domain with the characteristic cysteine of the CAAX motif (see Figure 6). Moreover, K-Ras is unique, as it is not palmitoylated within the hypervariable region. These characteristics prompted us to investigate the importance of the lysine residues within the farnesylated domain. Lysine K182 and K184 in K-Ras were mutated to the corresponding serine in H-Ras and the corresponding proline in N-Ras (also see Figure 6 for an overview). Co-precipitation assays, as illustrated in Figure 7A-C, showed that only exchanging one of these lysines did not significantly alter the binding of Galectin-8. However, replacing both lysines significantly reduced the interaction index of Galectin-8 long for K-Ras to 0.49 ± 0.37 (p = 0.038) and of Galectin-8 short to 0.56 ± 0.35 (p = 0.047) as compared to the wildtype K-Ras form ( Figure 7A). Introducing a lysine in position 183 of H-Ras, resulting in two lysine residues next to the farnesylated cysteine, increased the interaction index of both Galectin-8 isoforms for H-Ras by <1.5-fold, as compared to the wildtype form of H-Ras ( Figure 7B). However, exchanging K185 in H-Ras to proline did not alter the binding index, which indicated that this lysine alone is not critical for the interaction. N-Ras exhibit no interaction with Galectin-8, but introducing lysine residues in position 183 and 185 resulted in a detectable amount of co-precipitated Galectin-8 short. The insertion of K185 alone produced a weaker interaction with Galectin-8 short.

Removal of the Hexa-Lysine PBD Markedly Abrogates the K-Ras-Galectin-8 Interaction
EGFP-K-Ras (G12V) mutants were generated in which the six lysines in the PBD were replaced by the corresponding amino acids of either H-Ras (PL-H) or N-Ras (PL-N) to investigate whether the positively charged PBD (aa 175-180) contributes to the K-Ras-Galectin-8 interaction. Moreover, the lysines in the PBD and the two lysine residues (K182 and K184) in the farnesylated domain were both substituted (see Figure 6C for details). The mutants were expressed as EGFP-fusion proteins in the Galectin-8 negative HEK293 cells, immunoprecipitated, and then incubated with PANC-1 cell lysate to co-precipitate Galectin-8. As exemplified in Figure 8, the substitution of the hexa-lysine stretch in K-Ras caused a significant decrease in the interaction index to 0.27 ± 0.16 for Galectin-8 short and 0.14 ± 0.036 for Galectin-8 long in the case of PL-H, and 0.235 ± 0.12 and 0.16 ± 0.09 in the case the of PL-N when compared to K-Ras (G12V) set to 1.0. When all eight lysine residues in the C-terminus of K-Ras were exchanged, only a faint precipitation of the short isoform of Galectin-8 was evident and the interaction index dropped below 0.02. In summary, the specificity of the interaction of Galectin-8 with K-Ras is indispensably associated with the farnesylation of K-Ras and it is reliant on the existence of multiple lysines within the hypervariable region.

Removal of the Hexa-Lysine PBD Markedly Abrogates the K-Ras-Galectin-8 Interaction
EGFP-K-Ras (G12V) mutants were generated in which the six lysines in the PBD were replaced by the corresponding amino acids of either H-Ras (PL-H) or N-Ras (PL-N) to investigate whether the positively charged PBD (aa 175-180) contributes to the K-Ras-Galectin-8 interaction. Moreover, the lysines in the PBD and the two lysine residues (K182 and K184) in the farnesylated domain were both substituted (see Figure 6C for details). The mutants were expressed as EGFP-fusion proteins in the Galectin-8 negative HEK293 cells, immunoprecipitated, and then incubated with PANC-1 cell lysate to co-precipitate Galectin-8. As exemplified in Figure 8, the substitution of the hexa-lysine stretch in K-Ras caused a significant decrease in the interaction index to 0.27 ± 0.16 for Galectin-8 short and 0.14 ± 0.036 for Galectin-8 long in the case of PL-H, and 0.235 ± 0.12 and 0.16 ± 0.09 in the case the of PL-N when compared to K-Ras (G12V) set to 1.0. When all eight lysine residues in the C-terminus of K-Ras were exchanged, only a faint precipitation of the short isoform of Galectin-8 was evident and the interaction index dropped below 0.02. In summary, the specificity of the Besides Ras-isoforms, Rho GTPases exhibit a comparable C-terminal domain, although these proteins are geranylgeranylated instead of farnesylated. The comparison of the C-terminal amino acids of K-Ras with Rac, Rho, and Cdc42, as shown in Supplementary Figure S3A, demonstrated that Rac1, its splice variant Rac1b, and Cdc42 also possess two lysine residues in the prenylated domain next to the C-terminal cysteine. Two additional lysine residues are in the adjacent domain in the case of Rac1/Rac1b. RhoA and RhoC exhibit a glycine next to the CAAX motif instead. Precipitation analyses with EGFP-K-Ras (G12V) as a reference revealed that Galectin-8 showed a similar interaction to EGFP-tagged Rac1, Rac1b, and Cdc42. Almost no interaction was evident for EGFP-RhoA or -RhoC. Thus, Rac and Cdc42 may also interact with Galectin-8 in vivo, if they are not occupied by RhoGDIs.
next to the C-terminal cysteine. Two additional lysine residues are in the adjacent domain in the case of Rac1/Rac1b. RhoA and RhoC exhibit a glycine next to the CAAX motif instead. Precipitation analyses with EGFP-K-Ras (G12V) as a reference revealed that Galectin-8 showed a similar interaction to EGFP-tagged Rac1, Rac1b, and Cdc42. Almost no interaction was evident for EGFP-RhoA or -RhoC. Thus, Rac and Cdc42 may also interact with Galectin-8 in vivo, if they are not occupied by RhoGDIs.

Structural Model of the Galectin-8 Binding Interface
Galectin-1-like N-CRD and a Galectin-2-like C-CRD that are connected by a hinge region ( Figure 9A), whose length varies considerably within the human isoforms, comprise Galectin-8. The group of Y. Kloog identified a potential farnesyl binding pocket in Galectin-1 by superimposing the secondary structures of Galectin-1 and RhoGDI [39] using the Cdc42-RhoGDI complex as a model [40]. In RhoGDI, this pocket expands upon binding a geranylgeranyl group. We superimposed the N-terminal domain of Galectin-8 short with Galectin-1 and identified a hydrophobic channel in Galectin-8 that seems to be too narrow to bind a farnesyl chain to localize a putative farnesyl binding pocket in Galectin-8. However, there are two potential binding pockets in the N-terminal domain with the assumption that the binding-pocket expands upon binding the farnesyl group ( Figure 9A). Pocket A is similar to the farnesyl binding pocket that was proposed by Rotblat [39] and pocket B is located on the other side of the hydrophobic channel. Consequently, the farnesyl group would enter

Structural Model of the Galectin-8 Binding Interface
Galectin-1-like N-CRD and a Galectin-2-like C-CRD that are connected by a hinge region ( Figure 9A), whose length varies considerably within the human isoforms, comprise Galectin-8. The group of Y. Kloog identified a potential farnesyl binding pocket in Galectin-1 by superimposing the secondary structures of Galectin-1 and RhoGDI [39] using the Cdc42-RhoGDI complex as a model [40]. In RhoGDI, this pocket expands upon binding a geranylgeranyl group. We superimposed the N-terminal domain of Galectin-8 short with Galectin-1 and identified a hydrophobic channel in Galectin-8 that seems to be too narrow to bind a farnesyl chain to localize a putative farnesyl binding pocket in Galectin-8. However, there are two potential binding pockets in the N-terminal domain with the assumption that the binding-pocket expands upon binding the farnesyl group ( Figure 9A). Pocket A is similar to the farnesyl binding pocket that was proposed by Rotblat [39] and pocket B is located on the other side of the hydrophobic channel. Consequently, the farnesyl group would enter these potential binding pockets from different sides of the protein. They both proposed hydrophobic binding pockets have to expand upon the binding of the farnesyl molecule, but pocket A requires more structural changes than pocket B. A remarkable difference at the entrance of the pockets is the electrostatic potential. The surface charge of pocket B is negative in contrast to pocket A ( Figure 9B). Residues D25, D28, E133, and D136, and residues D156 and E166 from the linker domain form this cluster of negatively charged amino acids in pocket B ( Figure 9C). Our binding studies (summarized in Figure 6B) show that the farnesyl moiety of K-Ras and its polybasic region is indispensable for the complex formation with Galectin-8 via the N-CRD; therefore, we suggest that a farnesyl binding pocket in Galectin-8 must exist in conjunction with a negative cluster of amino acids that interact with the positively charged lysine stretch of K-Ras. A cluster of negatively charged residues is only found in the surrounding of the proposed binding pocket B. For these reasons, we assume that, in Galectin-8 pocket B ( Figure 9A) and not pocket A, as proposed for Galectin-1, binds the farnesyl moiety of K-Ras. Furthermore, the homologous region in the C-CRD is not negatively charged, which is consistent with our findings that C-CRD does not interact with K-Ras ( Figure 5). electrostatic potential. The surface charge of pocket B is negative in contrast to pocket A ( Figure 9B). Residues D25, D28, E133, and D136, and residues D156 and E166 from the linker domain form this cluster of negatively charged amino acids in pocket B ( Figure 9C). Our binding studies (summarized in Figure 6B) show that the farnesyl moiety of K-Ras and its polybasic region is indispensable for the complex formation with Galectin-8 via the N-CRD; therefore, we suggest that a farnesyl binding pocket in Galectin-8 must exist in conjunction with a negative cluster of amino acids that interact with the positively charged lysine stretch of K-Ras. A cluster of negatively charged residues is only found in the surrounding of the proposed binding pocket B. For these reasons, we assume that, in Galectin-8 pocket B ( Figure 9A) and not pocket A, as proposed for Galectin-1, binds the farnesyl moiety of K-Ras. Furthermore, the homologous region in the C-CRD is not negatively charged, which is consistent with our findings that C-CRD does not interact with K-Ras ( Figure 5). As stated above, the N-CRD resembles Galectin-1, thus we investigated whether EGFP-K-Ras (G12V) from PANC-1 cell lysates also co-precipitated Galectin-1. PANC-1 cells endogenously express Galectin-1, Galectin-3, and Galectin-8, but to considerably different amounts. Densitometric analysis of representative western blots using the amount of recombinant Galectin-1 and Galectin-8 short as references calculated the amount of Galectin-1 and Galectin-8 short in PANC-1 cell lysates. In PANC-1 cell lysates, approximately 0.05 µg/mg lysate Galectin-8 short and 1.7 µg/mg lysate As stated above, the N-CRD resembles Galectin-1, thus we investigated whether EGFP-K-Ras (G12V) from PANC-1 cell lysates also co-precipitated Galectin-1. PANC-1 cells endogenously express Galectin-1, Galectin-3, and Galectin-8, but to considerably different amounts. Densitometric analysis of representative western blots using the amount of recombinant Galectin-1 and Galectin-8 short as references calculated the amount of Galectin-1 and Galectin-8 short in PANC-1 cell lysates. In PANC-1 cell lysates, approximately 0.05 µg/mg lysate Galectin-8 short and 1.7 µg/mg lysate Galectin-1 is present, which is >30 times more Galectin-1 than Galectin-8 short (Supplementary Figure S3B). Western blot analyses of K-Ras (G12V) co-precipitates showed that Galectin-8 and Galectin-1 were both detected in nearly equal amounts in the same precipitate, as referred to the recombinant control protein. But, considering that much less Galectin-8 was present in the sample, calculation of the bound Galectins revealed that approximately 45% of Galectin-8 short, but only 1.3% of Galectin-1 was co-precipitated with K-Ras (G12V) (Supplementary Figure S3, left panel and bar graph in the middle). These results clearly support that K-Ras preferentially interacts with Galectin-8 and not with Galectin-1 when present in the same lysate. These findings support our assumption that the negatively charged region of Galectin-8 (Figure 9), which is missing in Galectin-1, is crucial for complex formation with K-Ras.
Exchanging the two lysine residues in position 182 and 184 of K-Ras nearly completely abrogated the binding of Galectin-1, whereas the exchanging of lysine 182 had no effect and the exchanging of lysine 184 downregulated binding to 50% when compared to the wildtype K-Ras (Supplementary Figure S3, right bar graph). This observation implies that the two C-terminal lysine residues are also important for the binding of K-Ras to Galectin-1, probably by hydrophobic interactions mediated by the long hydrophobic lysine side chains. However, K-Ras favors complex formation with Galectin-8. Therefore, we suppose that the hydrophobic effects of the two C-terminal lysine residues are not as crucial for Galectin-8 as for Galectin-1 binding due to the additional electrostatic interactions within the N-CRD of Galectin-8 with the multiple lysines in K-Ras.
In conclusion, we have demonstrated within this work that Galectin-8 is a new and direct binding partner for K-Ras. The N-CRD specifically mediates the interaction with the farnesylated PBD of K-Ras, thereby modulating K-Ras effector pathways as well as cell proliferation and migration. The current results underscore the potential of Galectin-8 as a new intracellular signaling modulator and pharmacological target in K-Ras-dependent tumors. Inhibiting Galectin-8 and, hence, interfering with K-Ras signaling, might represent a new strategy to block tumor cell proliferation and motility and probably metastatic dissemination.

Discussion
In the presented study, we have combined the biochemical, structural, and cellular analyses to identify and characterize Galectin-8 as a new direct binding protein for K-Ras, which modulates the signal transduction pathways downstream of Ras and cellular functions involved in tumorgenesis.
Our precipitation experiments demonstrate that the two isoforms of Galectin-8 directly bind to K-Ras. These two isoforms differ with regard to the expression of exon 11, encoding for 42 amino acids located in the hinge region between the N-and C-terminal CRD. Both isoforms are expressed at the protein level in 19 of 20 analyzed lung and pancreatic carcinoma cell lines (Supplementary Figure S1). This high frequency correlates with the findings of different groups, which showed that Galectin-8 is one of the most widely expressed tandem-repeat galectins in human cells and tumors [41,42]. Most of the studies concentrate on extracellular functions of secreted Galectin-8, which, for example, can either operate as a matrix protein, thereby supporting cell adhesion, or as a soluble ligand, thereby inhibiting cell adhesion by modulating integrin β1 signaling [20,[41][42][43]. However, its role in oncogenesis and tumor progression is not well understood, since its expression can be found as being increased or decreased in different tumor types [41,44]. The increase in the Galectin-8 gene (LGALS8) expression correlates with increased human lung cancer progression and metastasis [45]. Unpublished, preliminary studies from our group indicates increased Galectin-8 mRNA expression in over 50% of the analysed sample from pancreatic carcinoma patients as compared to normal tissue. However, Galectin-8 seems to be down-regulated during the progression of pancreatic cancer, as detected in another study using immunohistochemistry [46]. Therefore, the available data point towards an organ and tumor-specific function of  Galectins shuttle between the nucleus, cytoplasm, and extracellular compartment. Targeting to the membrane is especially evident for Galectin-1 and -3 when interacting with active H-Ras or K-Ras, respectively [21]. Cell lysate fractionation analyses, as shown for PANC-1 cells in this study ( Figure 2) and several other carcinoma cell lines, revealed that Galectin-8 is localized in the cytosolic and membrane-containing fraction and it co-localizes with K-Ras at the plasma membrane. Essential studies by Hancock and colleagues have established that farnesylated K-Ras is mainly localized in non-raft microdomains through the electrostatic interaction of its PBD with the plasma membrane and it is recruited into nanoclusters after GTP loading [7,14,47]. A complex interaction of the farnesyl group and the PBD with acidic phospholipids, such as phosphatidylserine, mediates nanocluster formation. Up to approximately 40% of the K-Ras molecules are arranged in nanoclusters of~6 proteins, whereas the remaining molecules are localized as monomers or dimers in the plasma membrane [14,48]. K-Ras nanocluster formation is modulated by PKC-mediated phosphorylation of S181, which changes the charge of the PBD of K-Ras [10], as well as by Galectin-3 [23]. Galectin-3 binds to membrane lipids, including phosphatidylserine, and sequesters the farnesyl moiety of K-Ras within in its hydrophobic pocket. This stabilizes K-Ras.GTP nanoclusters which act as a signaling hub to recruit downstream effectors, such as Raf-1 and PI3K [16,23]. The data presented here demonstrate that the interaction of K-Ras and Galectin-8 relies heavily on both the farnesyl moiety and the lysine residues within the HVR. K-Ras lacking farnesylation did not interact with Galectin-8, pointing towards a farnesyl binding pocket that is similar to the ones already discovered in Galectin-1, Galectin-3 [23,39,49], and PDEδ [11]. Sequence alignments and structural comparisons showed that Galectin-8 exhibits a Galectin-2-like C-CRD and a Galectin-1-like N-CRD (Figure 9), with the latter showing two potential farnesyl binding pockets. We propose that the N-CRD mediates binding to K-Ras via the hydrophobic pocket B, since the N-CRD alone or in conjunction with the hinge region strongly interacted with farnesylated K-Ras ( Figure 5). This pocket is surrounded by a cluster of negatively charged amino acid residues, which could interact with the positively charged lysine residues within the HVR of K-Ras, as depicted in Figure 9C. This negative cluster is formed by residues D25, D28, E133, and D136, which are expressed in all Galectin-8 variants, and supported by residue D156 and E166 from the hinge domain. There are x-ray structures of the N-and the C-terminal subunits of Galectin-8, but none for the hinge region, which might lead to certain inaccuracies in the proposed folding of the hinge domain. The superimposition of the N-and C-CRD of Galectin-8 shows that both domains share a similar fold. However, no negative cluster is found in the equivalent potential cavity region of the C-CRD. Thus, we assume that Galectin-8 binds only one K-Ras molecule via the N-CRD. Furthermore, a comparison of our model with other Galectin-8 structures (RCSB Protein Data Bank: 4han, 4fqz) and with those of Galectin-3 (4lbo) and Galectin-1 (3t2t) revealed a very similar overall fold but the described negative cluster is only present in Galectin-8. Our proposed model is in accordance with the experimental data, which showed that, besides the farnesylation, the two lysines adjacent to the farnesylated cysteine, as well as the six lysines in the PBD of K-Ras, are crucial for the interaction (as summarized in Figure 6B,C). Moreover, introducing a second lysine into the farnesylated domain of H-Ras enhances the weak binding to Galectin-8, which points towards a supportive electrostatic interaction with the negatively charged residues in the binding cavity of Galectin-8. However, we cannot entirely exclude that H-Ras uses pocket A on the opposite side of the molecule, instead of pocket B.
Until now, Galectin-3 has been discussed as a chaperone-like protein for K-Ras.GTP leading to nanoclustering and subsequent downstream signaling. The K-Ras-Galectin-3 complexes were originally discovered in NIH/3T3 cells [24], a cell line that hardly expresses Galectin-8. In our precipitation analyses with K-Ras, we could hardly detect co-precipitated Galectin-3. Thus, Galectin-8 might either compete with Galectin-3 for binding to K-Ras or it might interact with a subset of K-Ras molecules not targeted by Galectin-3. The finding supports the latter hypothesis that Galectin-8 interacts with GTP-, as well as GDP-bound K-Ras ( Figure 5C), whereas Galectin-3 only interacts with K-Ras.GTP trapped in nanoclusters [23]. Galectin-8 might target these molecules, given that 50-60% of the K-Ras molecules are segregated as monomers in the membrane, possibly in a latent signaling state. As demonstrated here, deletion of Galectin-8 led to an upregulation of the total and membrane-bound EGFP-K-Ras (G12V) protein content ( Figure 2C,E). This might be the result of the stabilization of GTP-bound EGFP-K-Ras (G12V) by binding now to Galectin-3 and by reduced cycling and/or proteasome-mediated degradation of K-Ras. The trafficking, recycling, and degradation of K-Ras is controversially discussed within the literature. K-Ras, phosphorylated via PKC on S181, traffics from the plasma membrane to the ER and Golgi [10], and it can be endocytosed and targeted to early and late endosomes, as well as lysosomes [35]. The inhibition of lysosomal degradation extended the half-life of K-Ras and caused sustained ERK activation after EGF stimulation [31,35]. The group of J. Hancock reported a mislocalization of K-Ras by using Fendiline and Staurosporine, thereby inhibiting the interaction with the plasma membrane and directing K-Ras to the cytosol, endosomes, mitochondria, Golgi, and the ER [50,51]. Galectin-8 might serve as an escort protein for K-Ras endosomal trafficking, since we observed Galectin-8 in the membrane of Rab-7 containing endosomes (our unpublished results). Polyubiquitination also controls K-Ras protein stability [36]. Our data, showing that the proteasome inhibitor MG132 increases the steady-state level of EGFP-K-Ras (G12V) similar to the effect seen after Galectin-8 depletion ( Figure 3B), might also offer the possibility that Galectin-8 is involved in proteasome-mediated K-Ras degradation. The enhanced EGFP-K-Ras (G12V) content in the analyzed PANC-1 cell clones (Figures 2 and 3) after Galectin-8 depletion resulted in enhanced ERK activation, which enhanced K-Ras (G12V) nanocluster formation could mediate by binding to Galectin-3 and it recruitment to the plasma membrane.
K-Ras traffics via a still uncharacterized Golgi-independent pathway to the plasma membrane. Thus, Galectin-8 might primarily act as a GDI-like solubilizing factor by shielding the farnesyl group and assisting its intracellular trafficking to or from membranes, as described for PDEδ [11,52]. PDEδ controls the accurate localization of K-Ras at the plasma membrane by enhancing the solubilization and transmembrane exchange of K-Ras from endomembranes to the plasma membrane. The depletion of PDEδ results in reduced plasma membrane localization and ERK signaling in MDCK cells [52]. Our data show that Galectin-8 depletion upregulates EGFP-K-Ras (G12V) protein content in the particulate and slightly in the soluble fraction ( Figure 2E). These data might point towards opposite functions of Galectin-8 and PDEδ in mediating plasma membrane to endomembrane trafficking and the localization of K-Ras. Further work is needed for solving the molecular mechanism how Galectin-8 acts as a regulator for the dynamic equilibrium of membrane-bound and cytosolic K-Ras and its protein stability.
It is well documented that human pancreatic and lung cancer with oncogenic mutation in the K-Ras allele is highly aggressive and is associated with poor prognoses. Fig. 4 documents a function for Galectin-8 in controlling tumor-relevant functions, such as cell proliferation and migration. Cell migration was reduced by approx. 30% and the doubling time increased by approx. 25% to 65%, depending on the cell clone, leading to a reduction in cell proliferation, upon the depletion of Galectin-8. The moderate reduction of cell proliferation and migration might be due to the considerable amount of Galectin-8 that is still present in the cells after siRA treatment. A CRISPR/Cas-mediated inhibition of Galectin-8 would be required to further validate Galectin-8 as a pharmacological target.
Taken together, we have identified Galectin-8 as a new and direct binding partner for K-Ras and have characterized the respective binding interfaces in Galectin-8 and K-Ras. The current results underscore the potential of Galectin-8 as a new intracellular signaling molecule and possible pharmacological target in K-Ras-dependent tumors since Galectin-8 modulates K-Ras-induced signaling, proliferation, and migration. Inhibiting Galectin-8 itself or its binding to K-Ras and, thus, impeding K-Ras signaling might represent a new strategy to block tumor cell proliferation and motility and probably metastatic dissemination.

Antibodies, Inhibitors and Recombinant Proteins
Antibodies, inhibitors, and recombinant proteins were obtained from standard suppliers and are listed in Supplementary Materials.

Protein Lysate Preparations
Cell lysates were prepared in Gold-Lysis buffer or for phosphorylation analyses in IP buffer that was supplemented with proteinase and phosphatase inhibitors, being homogenized by forcing the suspension through a 0.45 × 25 mm needle attached to a syringe 5-10 times, and cleared via centrifugation at 16100× g for 15 min. at 4 • C. Western blotting analyzed 50 µg of protein lysates.

Purification of His-Tagged Proteins Using Ni-NTA Agarose Resin
Baculovirus-infected H5 insect cells were maintained for 96 h at 27 • C in Ex-Cell 405 medium (Sigma-Aldrich, Germany) and then lysed in His buffer. Ni-NTA Agarose Resin (Serva, Heidelberg, Germany) was equilibrated with His buffer without 0.5% (m/V) sodium deoxycholate and incubated with cell lysate for 1 h at 4 • C by end over end rotation. The resin was washed 4 x with Ni-NTA washing buffer containing 20 mM imidazole for 15 min. at 4 • C by end-over-end rotation. His-tagged proteins were eluted 3× while using 100 µL of elution buffer [Ni-NTA washing buffer containing 250 mM imidazole] by end-over-end rotation for 10 min. at room temperature.

Ras Activity Assays
The GTP-bound forms of K-Ras, H-Ras, and N-Ras were recovered from cell lysates by affinity precipitation while using a GST fusion protein of the Ras-binding domain (RBD) of Raf-1 as an activation-specific probe for Ras-GTP essentially as described in [34].

Migration and Proliferation Assays
For wound healing assays, PANC-1 cells were seeded into six-well plates (8 × 10 5 cells/well), transfected with 50 nM siRNA, and cultured for 24 h, as described above. After treating the cells with mitomycin-C (10 µg/mL) for 2 h, three parallel scratches were made into the confluent monolayer while using a pipette tip. The washed cells were incubated in DMEM without supplements. Images of six randomly chosen areas at the scratches were taken every 3 h for 48 h using EVOS FL Auto 2 Cell Imaging System (ThermoFisher Scientific, Germany). Stacks of all images made for one point within the scratch were generated using Fiji (Version 1.52i) and analyzed by Scratch Assay Analyzer Plugin (MiToBo, Universität Halle) [55]. Cell migration was calculated in [µm/h]. The velocities were normalized to the control of each cell line. For proliferation assays, the cells were transfected with 33 nM siRNA. 24 h later, cells (5 × 10 4 ) were seeded into 12-well plates and then incubated for 12 h to allow for cell attachment. Images of 5-6 randomly chosen areas/well were taken in 12 h interval, using EVOS FL Auto 2 Cell Imaging System. The cells were counted while using Fiji software at time points 0 h (= 36 h after transfection) at least 48 h (= 84 h after transfection). The doubling time was calculated for each experimental setup.

Structural Modeling
The structures of two human Galectin-8 isoforms [Gal-8s: 317 aa, Gal-8l: 359 aa] were modeled according to the crystal structures from human Galectin-8 (4han: 293AS [56] and 4fqz: 291 aa [57]) while using the SWISS-MODEL automated comparative protein modeling server [58]. In the modeled range, the sequence identity of Gal-8s and Gal-8l to 4han and to 4fqz is at least 99.0%. Our model was superimposed with other structures via secondary structure matching (SSM) [59]. The farnesyl moiety, which was linked to the C-terminal end of K-Ras (taken from 5tar [60]), was manually modeled into the binding pocket with the interactive graphics program Coot [61], subsequently geometry minimization of the Galectin-8-geranylgeranyl complex was performed with PHENIX [62].

Miscellaneous
The bicinchoninic acid (BCA) assay determined the protein concentration (Pierce, Sankt Augustin, Germany). Ponceau S staining controlled the equal loading and transfer of proteins (Sigma-Aldrich, Taufkirchen, Germany). Immunoreactive bands were detected by using the appropriate antibodies and either ECL Western Blotting Detection System (GE Healthcare, Munich, Germany) and Fusion Imaging System and Software (Vilbert Lourmat, Eberhardzell, Germany) or Odyssey Infrared Imaging System and Software (LI-COR Biosciences, Bad Homburg vor der Höhe, Germany). All of the experiments were repeated as indicated, and similar results and identical trends were obtained each time. Data from representative experiments are shown. Western blot quantification was performed while using Image Studio Lite Software (Li-COR Biociences, Bad Homburg vor der Höhe, Germany). All of the values are presented as mean ± SD or SEM and statistical significance was calculated using one sample t test (Prism 7, GraphPad Software Inc., San Diego, CA, USA). A value of p ≤ 0.05 was considered to be significant.

Conclusions
In summary, our study uncovered Galectin-8 as a novel intracellular binding partner for K-Ras4B, which modifies K-Ras stability, ERK signaling, and migration and proliferation in pancreatic and lung carcinoma cells. The biochemical and structural characterization of the K-Ras-Galectin-8 interacting domains revealed that the binding is mediated via the farnesyl moiety of K-Ras and is highly supported by the polybasic stretch of lysine residues in the C-terminus of K-Ras. With regard to Galectin-8, the interaction is facilitated by the N-CRD and not the C-CRD. The N-CRD harbors a pocket for incorporating the farnesyl moiety of K-Ras, and a cluster of negatively charged amino acids, which presumably interacts with the positively charged lysines of K-Ras. The inhibition of Galectin-8 and, hence, interference with K-Ras signaling, may represent a new strategy for blocking tumor metastasis, since Galectin-8 is abundantly expressed in carcinoma.