The Animal Lectin Galectin-8 Promotes Cytokine Expression and Metastatic Tumor Growth in Mice

Secreted animal lectins of the galectin family are key players in cancer growth and metastasis. Here we show that galectin-8 (gal-8) induces the expression and secretion of cytokines and chemokines such as SDF-1 and MCP-1 in a number of cell types. This involves gal-8 binding to a uPAR/LRP1/integrin complex that activates JNK and the NFkB pathway. Cytokine and chemokine secretion, induced by gal-8, promotes migration of cancer cells toward cells treated with this lectin. Indeed, immune-competent gal-8 knockout (KO) mice express systemic lower levels of cytokines and chemokines while the opposite is true for gal-8 transgenic animals. Accordingly, gal-8 KO mice experience reduced tumor size and smaller and fewer metastatic lesions when injected with cancer cells. These results suggest the existence of a ‘vicious cycle’ whereby gal-8 secreted by the tumor microenvironment, promotes secretion of chemoattractants at the metastatic niche that promote further recruitment of tumor cells to that site. This study further implicate gal-8 in control of cancer progression and metastasis through its effects on the production of immunoregulatory cytokines.


Gal-8 induction of SDF-1 expression is independent of its sugar binding properties. Gal-8 acts as
an extracellular ligand that activates signaling pathways both by protein-sugar and protein-protein interactions 14 .
To determine how gal-8 triggers SDF-1/MCP-1 transcription, its sugar-binding activity was blocked by thiodigalactoside (TDG) 14 . Alternatively, we used a GST-gal-8 mutant, denoted W2Y, that lacks sugar-binding activity, due to mutation of two Trp residues (W85 and W248) crucial for sugar binding 14 into Tyr. Indeed, we could show that 20 mM TDG effectively inhibited the hemagglutination activity of gal-8, while the W2Y gal-8 mutant showed no such activity (Fig. 1h). We could further demonstrate (Fig. 1i) that TDG inhibited gal-8-mediated Akt (Protein kinase B (PKB)) phosphorylation, one of the hallmarks of gal-8's activity that involves protein-sugar interactions 23 . Similarly, unlike the naive gal-8, its W2Y mutant failed to stimulate Akt phosphorylation (Fig. 1j,k; Supplemental Fig. S2-a). These results indicated that indeed the W2Y gal-8 mutant lacks bioactivity which depends upon sugar binding. Next, the effects of TDG and those of GST-gal-8-W2Y on cytokine expression were evaluated. As shown in Fig. 1l, TDG partially inhibited RANKL expression induced by gal-8, while having no inhibitory effects on SDF-1/MCP-1 expression. Accordingly, the gal-8-W2Y mutant failed to induce RANKL expression while it was equipotent to naive gal-8 in promoting SDF-1/MCP-1 expression. These results indicate that gal-8's effects on SDF-1 and MCP-1 in osteoblast are independent of its sugar binding properties, whereas its effects on RANKL expression involve, at least in part, its sugar binding ability.

Gal-8 promotes chemoattraction of prostate cancer cells toward osteoblasts via SDF-1 and MCP-1.
Given that gal-8 induces expression and secretion of chemokines in different organs, we wished to determine whether it affects the migration of prostate cancer cell towards osteoblasts. Osteoblasts were seeded in one chamber of 'ibidi' culture-insert and were treated with or without gal-8 (50 nM, 24 h). Human prostate cancer (PC3) cells 35 were then seeded in the next 'ibidi' chamber. The 'ibidi' inserts were removed 24 h later and the PC3 cells were allowed to migrate for 6 h and close the gap between them and the osteoblasts. Treatment of osteoblasts with gal-8, prior to their interaction with the PC3 cells, increased ~2 fold PC3 cells migration towards these osteoblasts (Fig. 2a). Treatment of PC3 cells themselves with gal-8 did not affect their migration towards non-treated PC3 cells, while the osteoblasts themselves also failed to significantly migrate within the 6 h time frame of the experiments (Supplemental Fig. S3). These results suggest that gal-8 promotes chemoattraction of PC3 cells towards osteoblasts. Treatment of PC3 cells with Actinomycin-D that inhibits cell proliferation did not affect their accelerated migration towards gal-8-treated osteoblasts (Supplemental Fig. S4), indicating that the gap closure by PC3 cells is due to migration rather than extensive proliferation. Finally, PC3 cell migration towards gal-8-treated osteoblasts was recorded using live-cell imaging as described 36 . As shown in a snap-shot captured 6 h after initiation of the experiment (Fig. 2b), we observed accelerated migration of PC3 cells towards gal-8-treated osteoblasts, rather than PC3 cell proliferation.
The enhanced migration of PC3 cells (Fig. 2b) was indeed mediated by SDF-1 and MCP-1, secreted by gal-8-treated osteoblasts, because AMD3100, the SDF-1 receptor (CXCR4) inhibitor 37 , or the MCP-1 receptor Osteoblasts extracted from calvariae of newborn CD1 mice were incubated with or without 50 nM gal-8 for 24 h. Cells were harvested, total mRNA was extracted and qRT-PCR was conducted. HPRT served as a control for normalization purposes. Results shown are means ± SEM of 7-11 experiments done in duplicates. (b) Osteoblasts were transfected with pMET-Luc-SDF-1 vector. 24 h post transfection the cells were incubated in the presence or absence of 50nM gal-8 for 24 h, and luciferase levels in the conditioned medium were determined. Media from the above cells were used (c, d) to quantify the amounts of secreted SDF-1 (d) and MCP-1 (c) proteins. Results shown are means ± SEM of 3 experiments done in 5-8 repeats (b) or of two experiments done in triplicates (c,d). Liver (e), kidneys (f) and lungs (g) were removed from nine-week old CD1 mice. Single cell suspensions were made and were treated with 50nM gal-8 or serum-free medium (SFM; control) for 24 h. Cells were harvested, total mRNA was extracted and qRT-PCR was conducted for SDF-1 and MCP-1. HPRT served as a control for normalization purposes. Results shown (e-g) are means ± SEM of 3 experiments done in duplicates. (h) The ability of (CCR2)-antagonist 38 , either alone or in combination, effectively abolished the stimulatory effects of gal-8 on PC3 cell migration toward osteoblasts (Fig. 2c). Inclusion of specific SDF-1 antibodies but not control IgG, (Supplemental Fig S5) that prevents SDF-1 interactions with CXCR4 also reduced PC3 cells migration toward gal-8-treated osteoblasts (Fig. 2d). Gal-8-induced gap-closure, like its effects on cytokine secretion (Fig. 1l) were sugar-independent as inclusion of TDG failed to inhibit this process (Fig. 2e). These results conform with our hypothesis that gal-8 induces SDF-1 and MCP-1 secretion from osteoblasts and that those chemokines facilitate prostate cancer cell migration towards their target tissues. Of note, we observed certain variance in the percentage of gap closure in different experiments (e.g. Fig. 2c vs. 2e). This could be attributed to the fact that different batches of freshly isolated primary osteoblasts were used in the different experimental sets.

Gal-8 induction of SDF-1 and MCP-1 involves its binding to the low density lipoprotein receptor-related protein (LRP)-1 and the urokinase plasminogen activator receptor (uPAR) and activation of the c-Jun N-terminal kinase (JNK) and Nuclear factor kappa light-chain-enhancer of activated B cells (NFκB) signaling pathways.
We have previously shown that binding of gal-8 to MRC2/LRP1/uPAR receptor complexes in osteoblasts induces RANKL expression by these cells 16 . Therefore, we investigated the role of these receptors in mediating SDF-1 and MCP-1 secretion from osteoblasts treated with gal-8. Osteoblasts were transfected with small interfering RNAs (siRNAs) against LRP1, MRC2 or uPAR that silenced their expression by 80-90% each 16 (cf. also Supplemental Fig. S6). Next, the cells were treated with gal-8 (50 nM, 24 h). Silencing of uPAR and LRP1 resulted in ~50% decrease in SDF-1 transcription in response to gal-8, compared to cells treated with control siRNA. However, silencing of MRC2 did not impede the stimulatory effects of gal-8 on SDF-1 expression (Fig. 3a). These results suggest that the effects of gal-8 on SDF-1 expression in osteoblasts involve its binding to LRP1 and uPAR.κ To identify the signaling pathways mediating the effects of gal-8, primary osteoblasts were treated with MEK, Akt and p38 inhibitors, but these did not affect its ability to induce SDF-1/MCP-1 expression (Supplemental Fig. S7). In contrast, the JNK inhibitor SP600125 reduced SDF-1 and MCP-1 expression in response to gal-8 (Fig. 3b,c). Accordingly, and consistent with our previous studies in other cell types 23 , gal-8 induced phosphorylation (activation) of JNK in primary osteoblasts, and SP600125 inhibited this effect (Fig. 3d,e). These results suggest that JNK mediates the effects of gal-8, but additional signaling pathways are involved as well.
To explore the possible role of JNK as a downstream effector of LRP1 and uPAR, osteoblasts were transfected with LRP1 and uPAR siRNAs. 48 h post transfection the cells were further treated with SP600125 to inhibit JNK phosphorylation and then gal-8 was added for 24 h. We could show (Fig. 3f) that inhibition of JNK combined with silencing of uPAR had significant additive inhibitory effects on SDF1 transcription in response to gal-8, suggesting that uPAR signaling induced by gal-8 is not mediated by JNK. In contrast, the inhibitory effects due to silencing of LRP1 and JNK were not additive (Fig. 3f), suggesting that JNK is activated upon gal-8/LRP1 interactions.
Since MCP-1 and SDF-1 transcription is regulated by the (NFkB) 39,40 we analyzed the effects of gal-8 on this pathway in osteoblasts. Indeed, we could show a 2 fold increase in NFkB transcription activity in osteoblasts transfected with a pNFkB-MetLuc2-reporter that were treated with gal-8 for 4 h (Fig. 3g). This was accompanied by 3-4 fold increased phosphorylation (activation) of IKKα/β, the upstream activator of NFkB that was already evident after 3 min treatment of osteoblasts with gal-8 (Supplemental Fig. S8) and persisted for at least 30 min (Fig. 3h,i). A corresponding 40% reduction in protein levels of IkB, the downstream target of IKKβ, and the upstream activator of the NFkB pathway, was detected 30 min following treatment with gal-8 (Fig. 3j,k).
To establish NFkB as a mediator of gal-8's effects on SDF-1 and MCP-1 transcription, we silenced p100 and p105, the precursors of NFkB1 and NFkB2, respectively 41 . As shown in Fig. 3l the mRNA levels of p105 and p100 were reduced by 70-80% by their respective siRNAs. Furthermore, silencing of NFkB2 decreased the expression of SDF-1 and MCP-1, induced by gal-8, by 75%, and 65%, respectively, while silencing of NFκB1 had no such an effect (Fig. 3m,n). These results suggest that NFκB2 mediates, at least partially, SDF-1 and MCP-1 transcription in response to gal-8.

Alterations in cytokine/chemokine secretion in gal-8-transgenic (Tg) and knockout (KO)-mice.
To determine the physiological significance of the above findings we studied the levels of expression of cytokines/chemokines in gal-8 Tg and gal-8 KO mice. As expected, no expression of gal-8 mRNA itself was detected in osteoblasts isolated from newborn gal-8 KO mice ( Fig. 4a and 42 ). Furthermore, the mRNA levels of a number of cytokines and chemokines including MCP-1 and SDF-1 were significantly reduced (by 80-95%) in osteoblasts derived from these animals, when compared to wild-type (WT) mice (Fig. 4a). We also observed a 20 mM TDG to inhibit the hemagglutination activity of the indicated concentrations of gal-8 or its GST-W2Y gal-8 mutant were determined. Results shown are of a representative experiment repeated at least 6 times. (i,j) Primary osteoblasts were treated with GST-gal-8 in the absence or presence of 20 mM TDG (i), or with GST-W2YGal8 (50 nM) for 30 min (j). Cells were harvested, total proteins were extracted and were analyzed by Western blotting using antibodies specific for the phosphorylated forms of AKT (pAKT) or total general AKT (gAKT). A representative experiment is shown in (i,j) and quantification of three similar experiments is shown in (k). (l) Primary osteoblasts were treated for 24 h with gal-8 (50 nM) or GST-W2YGal-8 (50 nM), with or without TDG (20 mM). Cells were harvested, total mRNA was extracted and qRT-PCR was conducted to quantify the mRNA levels of SDF-1, MCP-1 and RANKL. Actin mRNA served as a control for normalization purposes. Results are mean ± SEM of 3 experiments done in duplicates (*p < 0.05; **p < 0.01; ***p < 0.001 vs. untreated controls). The full-length blots of (i,j) are shown in Supplemental Fig S10. Scientific RepoRtS | (2020) 10:7375 | https://doi.org/10.1038/s41598-020-64371-z www.nature.com/scientificreports www.nature.com/scientificreports/ 95% decrease in mRNA levels of inducible nitric oxide synthase (iNOS) that can be activated by pro-inflammatory cytokines.
Gal-8 Tg mice presented a mirror image to that of gal-8 KO mice. The mRNA levels of a number of cytokines (i.e. MCP-1, SDF-1, IP-10, IL-6, IL-1β, TNF-α), in addition to RANKL 16 , were increased in long bones of 14-15 weeks old mice, when compared to WT mice ( Fig. 4b), while the opposite was true for gal-8 KO mice. These results establish the role of gal-8 as a physiological regulator of cytokine/chemokine expression. To determine whether the reduced expression of cytokines/chemokines in gal-8 KO mice is indeed a systemic effect, mRNA was extracted from lungs and spleens of 7-weeks old mice. As expected, gal-8 KO mice did not express gal-8 mRNA in these tissues while the mRNA levels of IL-6, SDF-1, and MCP-1 were decreased 2-4 fold when compared to their WT controls (Fig. 4c,d). These results further establish gal-8 as a physiological systemic regulator of cytokine and chemokine expression in different tissues and cell types.
Gal-8 KO mice express lower levels of MMP9 and Gas6. Cytokines such as SDF-1 up regulate gene expression of MMPs 43 that play key roles in promoting cancer metastasis 44,45 . Therefore, we aimed to determine whether the mRNA levels of MMP9 are altered in gal-8 KO mice. Using RNA extracted from long bones of Gal8-KO mice we found significantly lower (50%) mRNA levels of MMP9 in gal-8 KO mice when compared to WT mice (Fig. 4e), suggesting that this might also contribute to the resistance of Gal-8 KO mice to develop cancer metastasis.

Gal-8 promotes cancer growth and metastasis in-vivo.
Given that cytokines and chemokines play key roles in tumor progression in vivo 33 and given that gal-8 promotes cytokine and chemokine expression in mice, we studied the effects of its depletion on cancer growth and metastasis. The proper model to study should have been the bone metastatic niche, however such a model is difficult to establish in immune competent C57Bl/6 J mice. Therefore, other models were employed. E0771 breast cancer cells 48 and D122-Luc Lewis lung carcinoma cells 49 , derived from C57BL6/J mice, that share isogeneic background with gal-8 KO mice were used. As shown in Fig. 5a-c (and Supplemental Fig. S9a,b) orthotopic injection of E0771 cells to the 4 th mammary gland of gal-8 KO female mice resulted in development of significantly lighter (~40%) primary tumors, having smaller size and volume than tumors that grew in WT mice. Similarly, injection of D122-Luc cells to the tail vein of gal-8 KO male mice resulted in development of lung metastatic lesion, having reduced weight (Fig. 5d) and fewer lesions (Fig. 5e) when compared to metastatic lesions developed in their WT control littermates (cf. also Supplemental  Fig. S9c). These results suggest that the lower levels of cytokines/chemokines expressed in gal-8 KO mice may contribute to the reduced formation of primary tumors and metastatic lesions in these animals.

Discussion
The present study employes cultured cells and immune competent mouse models to suggest the existence of a 'vicious cycle' whereby gal-8, secreted by tumor cells and their microenvironment, induces cytokine and chemokine production that supports further growth of primary tumors and metastatic lesions. Expression of additional mediators of tumor growth, including MMP9 and Gas6 is also induced by gal-8. The underlying mechanism involves binding of gal-8 to a complex of cell surface receptors that include LRP1 and uPAR; activation of the JNK and NFkB signaling pathways and induction of cytokine/chemokine production (e.g. MCP-1 and SDF-1). www.nature.com/scientificreports www.nature.com/scientificreports/ It occurs in a cell-autonomous manner in a number of cell types and tissues, suggesting that gal-8 induces a systemic response in vivo. This mechanism might be of clinical relevance, as the levels of expression of gal-8 at least in human bones, positively correlate with those of RANKL 16 and SDF-1 (unpublished).
Cytokines and chemokines are well known chemo-attractants that stimulate migration of malignant cells towards their metastatic niche 33 . Chemokine receptors are expressed by different cancer cells 50 and up-regulation of chemokine-receptor pairs (e.g. [Stromal cell-derived factor 1 (SDF-1)/ C-X-C chemokine receptor type 4 (CXCR4)] promotes metastasis 33 . CXCR4 enhances metastatic growth of breast cancer cells to bone, liver, and lung, tissues in which its ligand SDF-1 is expressed in high amounts 51,52 . SDF-1/CXCR4 signaling also benefits cancer cells by elevating the cells ability to express and secrete matrix metalloproteinases (MMPs) such as MMP9 53 . Similarly, Monocyte chemoattractant protein-1 (MCP-1) also known as CCL2 (C-C motif chemokine ligand 2), and its receptor CCR2 (C-C chemokine receptor type 2) were linked to tumor growth and poor prognosis in colon, breast, prostate, and cervical cancer 54 .
Certain effects of galectins 7,12 , including gal-8, on immune regulatory cancer networks were explored.   The present work provides a deeper insight into the molecular basis underlying the mode of action of gal-8 and highlights its physiological relevance. We show that engagement of the LRP1 and uPAR receptors by gal-8 triggers a signaling cascade that involves activation of JNK and NFkB2, leading to increased SDF-1 and MCP-1 expression. Using a combination of inhibitors and siRNAs, we further show that uPAR signaling is not exclusively mediated by JNK, while the interactions between gal-8 and LRP1 is mediated through JNK.
Being an animal lectin, a large number of its biological activities depend upon the sugar-binding capacity of gal-8 9 . Accordingly, mutations of key residues of gal-8 involved in sugar binding, abrogate several of its biological activities, including promotion of cell adhesion and spreading 14 . Here we show that expression of MCP-1/SDF-1 and chemoattraction of cancer cells are induced by gal-8 mostly independent of its sugar-binding activity. They are not blocked by inclusion of a sugar inhibitor (TDG), while being induced by a gal-8 mutant (gal-8 W2Y) that lacks a sugar binding activity 14 . These results indicate that gal-8 binding to its cell-surface receptors (e.g. LRP1/ uPAR) that triggers expression of MCP-1 and SDF-1 involves protein-protein, rather than protein-sugar interactions. In contrast, the effects of gal-8 on the expression of RANKL depend upon sugar-binding and involve interactions with MRC2 16 . Given that LRP1 is a negative regulator of RANKL expression 16 , while it is a positive regulator of SDF-1 expression, it is plausible that binding of gal-8 to LRP1, in a sugar-independent manner, negatively inhibits binding of gal-8 through its sugar-binding site, and therefore inhibits RANKL expression. The ability of gal-8, like other galectins, to engage in protein-protein interactions is well established [57][58][59] . Protein-protein interactions mediate gal-8 binding to NDP52, the autophagy cargo receptor 60 , while protein-protein interactions constitute part of the cytostatic effects of gal-1 61 .
The effects of gal-8 on cytokine/chemokine expression seem to have a physiological significance because total-body gal-8 KO mice 42 show reduced expression of cytokines and chemokines, including TNF-α, RANKL, IL-1β, SDF-1, MCP-1, IL-6, and IP-10 while the opposite is true for gal-8 Tg mice 16 that overexpress this lectin. The systemic reduction in cytokines and chemokines expression renders gal-8 KO animals partially resistant to growth and development of primary tumors and metastatic lesions. This is in accord with the notion that cytokines and chemokines promote growth of primary tumors, and support recruitment of cancer cells to the metastatic niche 33 . Hence, the action of gal-8 is most likely indirect and stems from its effects on the tumor microenvironment that secrets cytokines/chemokines in response to gal-8. This could account for the reduced size of the primary tumors, grown in gal-8 KO mice. This is also in accord with the observation that gal-8 does not control, in a cell-autonomous manner, primary growth of prostate cancer cells 28 .
Secreted MMP9 is a promoter of cancer metastasis. It acts by degrading the cellular matrix thus, supporting tumor cell invasion and spreading 62 . Here we show that gal-8 KO mice manifest reduced expression of MMP9 that could contribute to their resistance to tumor development. Gal-8 KO mice also have reduced expression of iNOS that can be activated by pro-inflammatory cytokines and may take part in anti-microbial activities 63 . While iNOS is commonly considered as being anti-tumorigenic, NOS2 has been implicated as a contributor to the process of tumor initiation and/or development, especially when the concentrations of its product, NO, are low. Therefore, inhibition of iNOS is proposed as a potential therapy in cases of triple-negative breast cancer 63 .
The third promoter of tumor growth whose expression is induced by gal-8 is Gas6, the ligand of the TAM family (Tyro3, Axl, and Mer) of receptor tyrosine kinases. Gas6 supports the development of several cancer types, (e.g. acute myelocytic leukemia, oral, prostate, renal, pancreatic, and ovarian cancers) 64 and its expression is associated with poor prognosis 65 . The ability of gal-8 to promote expression of Gas6, and the observed reduced expression of Gas6 in gal-KO mice, points at gal-8 as a potential physiological inducer of Gas6 expression that supports tumor progression.
Additional mechanisms may contribute to the pro-metastatic action of gal-8. These include promotion of homotypic aggregation of the tumor cells as well as increased cell-matrix interactions that increase cell growth, adhesion, and selective metastatic seeding 28,30,31 . This can be attributed to the role of gal-8 as an extracellular matrix protein, equipotent to fibronectin in promoting cell adhesion, spreading and migration 20 . Accordingly, gal-8 silencing inhibits filopodia formation, and aggregation of cancer cells 28 ; processes that are actively engaged in metastatic progression. with LRP1 or uPAR siRNAs (or control non-targeting siRNA). After 48 h SP600125 was added for 2 h and then gal-8 (50 nM) was added for additional 24 h, after which the cells were harvested, total mRNA was extracted and qRT-PCR was conducted in order to quantify changes in SDF-1 mRNA. HPRT served as a control. Results shown are means ± SEM of 3-4 experiments done in duplicates. (g) Osteoblasts (3 × 10 3 cells) were transfected with p-MET-LUC-NFkB construct. 24 h post transfection, cells were treated with gal-8 (50 nM) for 4 h. Conditioned medium was collected and luciferase secretion was determined. (h-k) Osteoblasts (10 5 cells) were treated with gal-8 (50 nM) for 30 min (h,i,k) or for the indicated time periods (j). At the end of incubation total proteins were extracted and analyzed by Western blotting using antibodies specific for the phosphorylated forms of IKKα/β (h,i) or IkB (j,k). Anti-general IKK (g-IKK) (h) or vinculin (j) served as controls. Results of 3 experiments done in duplicates were quantified (i,k). (l-n) Osteoblasts (5×10 4 cells) were transfected with NFkB1 (p105), NFkB2 (p100) siRNAs, or with control non-targeting siRNA. 48 h thereafter gal-8 (50 nM) was added for additional 24 h, after which the cells were harvested, total mRNA was extracted and qRT-PCR was conducted in order to quantify changes in mRNA levels of NFkB1 and NFkB2 (l) SDF-1 (m) or MCP-1 (n). Actin mRNA served as a control for normalization purposes. Results shown are means ± SEM of 3 experiments done in duplicates. (*p < 0.05; **p < 0.01; ***p < 0.001 vs. untreated controls). The full-length blots of (d,h,j) are presented as Supplemental Figs. S11-S13, respectively. Vertical black lines (j) represent positions where the original blots were cropped and non-relevant lanes were omitted.

Methods
General. All methods were performed in accordance with the relevant guidelines and regulations in effect at the Weizmann Institute of Science, Rehovot, Israel, and were approved by the institute.
Agglutination activity of gal-8. Hemagglutination activity was measured by mixing serial dilutions of gal-8 in PBS (50 μL/ well) with packed rabbit erythrocytes in PBS (50 μL/ well) in micro-titer U-shape plates. Following 1 h incubation at 22 °C, the agglutination activity was determined as described 19 .
Animals. Gal-8 KO mice and their wild-type controls (denoted WT1), as well as Gal-8 Tg mice and their wild-type controls (denoted WT2) were generated as we previously described 16,42 . Gal-8 KO mice underwent 7 backcrosses to C57Bl/6 J mice. The WT2 animals were CB6F1 mice as described 16 . All animals were housed under standard light/dark conditions in the animal care unit of the Weizmann Institute of Science. Mice were given food and water ad libitum. Experiments were approved by the Animal Care and Use Committee of the Weizmann Institute of Science. cell lines. E0771 breast cancer cells derived from C57BL mice were purchased from CH3 BioSystems (Amherst, NY). D122-Luc murine lung cancer cells expressing luciferase, were kindly provided by Prof. Lea Eisenach (Weizmann Institute of Science). Cells were grown according to the ATCC instructions. Primary osteoblasts from calvariae of newborn mice and long bones of adult mice were isolated as we described 16 . primary tumor models. 5 ×10 5 E0771 cells were injected into the 4 th mammary gland of 9-weeks old C57BL/6 J female mice. Mice were sacrificed 6 weeks post injection and the developed tumors were weighted and photographed.
Lung Metastasis models. 5 ×10 5 D122 cells were injected into the tail-vein of C57BL/6 J male mice. Six weeks post injection mice were sacrificed and the lungs were extracted, weighted, and examined for the presence of cancerous lesion.
RnA analysis. Cells were grown in 6-or 12-well plates. Following treatment, cells were harvested and total RNA was extracted using the PerfectPure RNA kit (5-prime). RNA was quantified and cDNA was generated by Figure 6. The 'vicious cycle' induced by Gal-8 to promote tumor growth and metastasis. Gal-8 drives cancer growth and metastasis by at least two mechanisms: i. Promotion of cell-matrix interactions that increase selective metastatic seeding through Gal-8 binding to α 3 β 1 and α 6 β 1 integrins, expressed by metastatic cells, that bind to Gal-8/fibronectin complexes at the metastatic niche and induction of MMPs; ii. Dissemination of gal-8, expressed by the primary tumor cells and by the tumor microenvironment that induces in an autocrine and paracrine manner the expression and secretion of cytokines and chemokines at the primary tumor site that promotes primary tumor growth. In addition, gal-8 secreted at the metastatic niche further enhances the production of cytokines/chemokines that chemoattract cancer cells to this site.