Galectin-9 signaling drives breast cancer invasion through matrix

Aberration in expression and function of glycans and their binding proteins (lectins) in transformed cells constitutes one of the earliest discovered hallmarks of cancer. Galectins are a conserved family of lectins that can bind to β-galactosides. Among them, the role of Galectin-9, a galectin with two carbohydrate binding domains in immune-tumor cell interactions has been well-established, although its effect on cancer cell behavior remains as yet unclear. In this study, we used a spectrum of cell lines from homeostatic breast cells to transformed non-invasive and invasive cell lines cultured in microenvironment-diverse conditions to show that Galectin-9 expression shows an elevation in association with invasiveness of breast cancer epithelia. Our observations were supported by immunohistochemical studies of breast tumors and adjacent normal-tissues from patients. Genetic perturbation of Galectin-9 as well as the pharmacological inhibition of activity using cognate inhibitors confirmed a positive correlation between Galectin-9 levels and the adhesion of the aggressive triple negative breast cancer cells MDA-MB-231 to- and their invasion through-extracellular matrices (ECM). Within a constituted organomimetic multiECM microenvironment, Galectin-9 enhanced both the solitary and the collective invasion of cancer cells. Quantitative proteomics led us to uncover the inductive role of Galectin-9 in the expression of the proinvasive protein S100A4. In addition, Galectin-9 expression correlated with FAK signaling, the inhibition of which decreased S100A4 mRNA levels. Our results provide crucial signaling insights into how the elevation in Galectin-9 expression in breast cancer cells potentiates their invasiveness through ECM during early steps of metastasis.

Widely expressed across tissues, galectins are seen both inside cells and in the extracellular space, and based on their location, modulate a diverse set of functions including, but not limited to, cell proliferation, apoptosis, pre-mRNA splicing, cell-cell and cell-ECM (extracellular matrix) adhesion, epithelial cell polarity, innate/adaptive immunity regulation, and migration (3)(4)(5)(6)(7). Given that these phenomena contribute to tissue architecture and homeostasis, aberrant expression of galectins have been increasingly found to be associated with carcinogenesis. In fact, galectins have been shown to regulate tumor cell growth, apoptosis, invasion, and metastasis, oncogenic signaling and immune suppression (3,8,9).
Galectin-9 belongs to a subset of the 15 mammalian galectins, with two different CRDs in one peptide chain, here termed N-CRD and C-CRD (10). Rigorous investigations on its functions have predominantly been confined to immunological contexts, such as the activation of dendritic cells and their induction of naïve T cell proliferation, a level dependent modulation of activated T cell fate and the regulation of B-cell driven autoimmunity (11)(12)(13) .
The role of Galectin-9 in breast cancer progression is as yet unclear. Although early studies showed an association of its expression with better prognosis in breast cancer patients, an increasing set of investigations implicate Galectin-9 as an invasion-promoting protein through its role in mediating the escape of cancer cells from immune surveillance. Recent evidence shows that the expression of Galectin-9 in cancers of breast, colon and in myeloid leukemia is regulated by proinvasive microenvironmental cues such as HIF-1 and TGF (14).
Overexpression of Galectin-9 in breast cancer has also been reported by Grosset and coworkers, where the subcellular localization of the protein was found to affect the prognosis of patients (15). In the context of the well-known demonstration of localization-specific roles of other galectins, and isoform-diverse expression of Galectin-9, a careful investigation of the effect of the latter on transformed cell phenotype is clearly warranted (15)(16)(17).
In this paper, we probed the expression of Galectin-9 within sections of patients with different breast cancer histological subtypes as well as an invasion-diverse set of breast cancer cell lines cultured on organotypic extracellular matrix (ECM) scaffolds. We show that the levels of Galectin-9 correlate with the ability of breast cancer epithelia to adhere to ECM as well as invade both collectively and as dispersed mesenchymal cells within reconstructed organotypic ECM-complex microenvironments. The potentiation of invasiveness is dependent on its upregulation of FAK signaling, which in turn regulates the metastasis-associated protein S100A4.

Results:
Galectin-9 is highly expressed in breast cancer epithelia Galectin-9 gene expression was measured in the cells cultured as monolayers ( Figure 1A) and on two ECM scaffolds using qRT-PCR: the nonfibrillar lrECM to mimic the milieu of cells in homeostasis or just before the basement membrane (BM) is breached ( Figure 1B) and the fibrillar Collagen I (Coll I), the stromal ECM milieu through which cancer cells invade ( Figure   1C). Galectin-9 transcript levels were significantly higher in MDA-MB-231 cultured on   plastic, lrECM and Coll I and in BT-549 cultured on plastic and Coll I compared with MCF7 and HMLE cells, which showed relatively lower levels of Galectin-9 mRNA ( Figure 1A-C and Figure S1A). The upregulation of Galectin-9 mRNA in MDA-MB-231 and BT-549 was consistent with higher levels of proteins in these cells as observed through immunoblotting, whereas HMLE and MCF7 showed sparse signals ( Figure 1D). To investigate if Galectin-9 levels were elevated in breast tumors in vivo, we stained histological sections from a cohort of 11 breast cancer patients along with their matched non-malignant tissues. Primary antibodyomitted controls were used to confirm the veracity of staining ( Figure S1B). Out of 11, 10 samples showed increased Galectin-9 levels in tumors compared with matched adjacent normal tissues (across all histotypes except in HER2 + subtype where Galectin-9 staining was elevated in tumors, but matched non-cancerous tissue was not available) ( Figure 1E-G & S1C). Our results were further confirmed by observation of an increased Galectin-9 mRNA and protein levels within breast tumors, when examined within publicly available samples curated by The Cancer Genome Consortium (TCGA; using RNA sequencing) and the Clinical Proteomic Tumor Analysis Consortium (CPTAC using a combination of large-scale proteome and genome analysis) respectively ( Figure S1D and E). In fact, its mRNA levels were found to be higher in breast cancer samples from stages 1,2 and 3 compared with adjacent non-cancerous control samples ( Figure S1F). Among the histopathological types, the most prevalent type: invasive ductal carcinoma (IDC) samples showed widely variable but significantly higher Galectin-9 mRNA and protein levels compared with noncancerous controls within the abovementioned databases (Figure S1G & S1H). Interestingly, within the TCGA dataset, we observed a significant hypomethylation of LGALS9 promoter within primary tumor samples compared with controls, suggesting a possible epigenetic upregulation of Galectin-9 associated with breast cancer progression (18,19) (Figure S1I).

Decrease in Galectin-9 expression and function decreases adhesion of breast cancer cells to ECM and their invasion through pathotypic ECM microenvironments
Having established that Galectin-9 levels were elevated in breast cancer cells, we sought to stably knock down its expression in MDA-MB-231 cells using a lentiviral transduction of genecognate shRNA (scrambled non-specific shRNA denoted in figures as shSc was used as control). Knockdown was confirmed using qRT-PCR and immunoblot ( Figure S2A & S2B).
Galectin-9-depleted cancer cells were observed to invade to a lower extent through lrECMcoated transwells, when compared with control cells (Figure 2A). In an earlier work, we have observed that invasion through ECM for MDA-MB-231 cells directly correlates with its ability to adhere to ECM substrata (20). Therefore, we assayed for, and observed that, the adhesion of cells to lrECM-and Coll I-coated substrata was significantly reduced upon depletion of Invasive breast cancer epithelia are capable of migrating within ECM using a combination of distinct modes: collective and mesenchymal single-cell invasion (21,22). This is evident when the cells are topologically surrounded by an outwardly radial ECM arrangement of BM and then Coll I, similar to what is seen in vivo (23). Accordingly, cancer cell clusters were prepared using Galectin-9-depleted and control cells, and after coating with lrECM, embedded in Coll I scaffolds as described earlier (see also Figure 2D top). It was evident from maximum intensity projections of confocal micrographs taken after 24 hours, that Galectin-9 knockdown significantly impaired cancer cell invasion into Coll I scaffolds ( Figure 2D bottom right) compared to control cells ( Figure 2D bottom left). A time-lapse video microscopic examination also revealed slower and sparser egress of breast cancer cells from the cluster into the surrounding collagenous microenvironment ( Figure 2E).
We sought to validate our findings in the context of the knockdown by designing and evaluating Galectin-9 specific inhibitors that impair its binding to β-galactosides. The design of novel more selective and potent Galectin-9 inhibitors was inspired by the recent publications that halophenyl thio--D-galactopyranosides show significant affinity enhancement over simple galactosides for most galectins (24) and that galactosides carrying a N-sulfonylamidine moiety at O3 display increased affinity and high selectivity for the N-terminal domain of Galectin-9 (25). Synthesis of two novel 3,4-dichlorophenyl thio--D-galactopyranosides carrying different N-sulfonylamidine moieties at O3, 2 and 3, were performed as reported (25)  Evaluation of galectin affinities were performed as reported (25,27) and clearly revealed both compounds being selective inhibitors of galectin-9 N-CRD with low µM affinities (Table 1).
Compound 2 is somewhat more potent than 3, while both show good, but to some extent different selectivity for Galectin-9 N-CRD over other galectins. Hence, 2 and 3 are suitable small-molecule inhibitors for evaluating effects of pharmacological inhibition of Galectin-9 in cell assays.   Similar with Galectin-9 knockdown, inhibition with 2 and 3 abrogated cancer cell migration into Coll I compared with vehicle-treated cells ( Figure 2J). Our genetic perturbation and small molecule inhibitor experiments suggest that Galectin-9 might play an inductive role in early cancer cell invasion.

Galectin-9 overexpression increases cancer cell adhesion and invasion through matrix
To understand if increased Galectin-9 positively correlates with invasive potential of cancer cells, we overexpressed its cDNA in MDA-MB-231 cells and confirmed its increased expression ( Fig S3). Overexpression of Galectin-9 led to an increase in the invasion of MDA-MB-231 cells through lrECM-coated transwells ( Figure 3A). Next, we assayed for adhesion of cancer cells to lrECM and Coll I. Overexpression resulted in increased adhesion of cancer epithelia to both the extracellular matrices ( Figure  Next, we asked if the increased cancer cell invasion shown by Galectin-9 overexpressing cells could be reversed by a cognate pharmacological impairment? To address this, we used the inhibitor 3 and allowed cancer cells to invade through multi-ECM scaffolds. Inhibition of elevated Galectin-9 levels decreased cancer epithelial invasion when compared to vehicle-treated Galectin-9 overexpressing cells ( Figure 3J). The reduced invasion upon inhibitor treatment was comparable to the vector control cells in Figure 3D. Our observations thus confirmed that Galectin-9 levels positively correlated with the invasive potential of cancer cells and might play an inductive role during breast cancer cell invasion. In addition, the pharmacological treatments suggested that such a correlation was dependent on the carbohydrate-binding activity of Galectin-9.

Galectin-9 potentiates cancer cell invasion by up-regulating S100A4 and focal adhesion kinase activation
We next sought to understand how Galectin-9 regulated cell adhesion to lrECM and Coll I, and invasion through multi-ECM environments. A quantitative proteomics approach was taken to identify the proteins that are dysregulated when Galectin-9 was overexpressed. We observed that 123 proteins were differentially expressed to significant extents (either upregulated or downregulated by 1.5-fold) ( Figure 4A). Functional annotation using ontological analysis showed such proteins to be significantly involved in cell-cell adhesion, cytoskeleton organization, and intermediate filament organization (Figure 4B). Similarly, KEGG pathway analysis also predicted that the differentially regulated proteins are involved in tight junction, actin organization and transendothelial migration pathways ( Figure 4C). We further probed proteins amongst top 20 upregulated proteins for validation using gene expression ( Figure S4).
Amongst the genes we analyzed, the expression of S100A4, a calcium binding protein was positively correlated with Galectin-9 levels ( Figure 4D and 4E). In addition, inhibition of Galectin-9 using the inhibitors 2 and 3 resulted in significantly reduced S100A4 gene expression ( Figure 4F).
During cancer cell invasion, cell-ECM adhesion has been shown to play a prominent role in mediating mesenchymal mode of invasion (28,29). Integrins are one of the major players in mediating cell adhesion to ECM. Outside-in signaling through integrins allows the phosphorylation of focal adhesion kinase (FAK) which in turn regulates actin polymerization dynamics that play a crucial role in cancer cell invasion (28,30). Therefore, we sought to probe whether FAK activation could be modulated by Galectin-9 levels.
Next, we probed for kinetics of pFAK upon inhibiting Galectin-9 using 3. We did not observe any immediate changes in pFAK levels at an early time point (30 min); at later time points, 1 hour and 6 hour, pFAK levels reduced when compared to 0 hour ( Figure 5C). These observations suggest that Galectin-9 regulates S100A4 expression and phosphorylation of FAK that are instrumental to the modulation of cancer cell adhesion and invasion. We next sought to verify a potentially inductive effect of FAK signaling on S100A4 as observed in other studies (31,32): in Galectin-9 overexpressing cells, which were treated with a pharmacological inhibitor of FAK signaling (CAS 4506-66-5; 100 nM), we observed a moderate downregulation in S100A4 mRNA compared with vehicle controls ( Figure 5D). These observations suggest that the upregulation of the migration-potentiating S100A4 by Galectin-9 is regulated by FAK signaling.

Discussion
An earlier study on Galectin-9 in breast cancer observed that its levels positively correlate with cell aggregation, thereby decreasing dissemination (27). However, the cell line used in the study, MCF7 invades sparsely, if at all. In addition, RNAseq-and immunohistochemistrybased studies indicate an increase in Galectin-9 mRNA and protein levels at advanced stages of cancer progression (18). In the present study, we rigorously confirmed the increase in Galectin-9 both in breast cancer patient samples using immunohistochemistry as well as in an invasion-diverse set of established cancer (and untransformed) cell lines. In addition, our experiments unambiguously demonstrate a positive correlation between Galectin-9 levels and invasiveness of cells through ECM. Our results seek to also reconcile the cell-aggregative behavior of Galectin-9 by showing its role in spurring bulk collective (radial) invasion in addition to single cell mesenchymal migration.
The inhibition of invasion phenotypes with Galectin-9 inhibitors with a strong binding affinity for N-CRD (and a relatively weak affinity for C-CRD) implies a relatively more prominent role of the former domain in regulating cancer invasion, but not exclude a role for the latter. Based on exon-intron organization, galectin CRDs have been phylogenetically divided into two types: F4 and F3 CRDs (10). The galectins, whose roles have been well elucidated in a variety of cancers: the prototype Galectin-1 and the chimeric Galectin-3, both possess F3 CRD domain (10). In contrast, the N-CRD of Galectin-9 is an F4 CRD. Along with reports suggesting F4 CRD of Galectin-9 being playing protumorigenic roles (33,34), our study showcases how structurally dissimilar galectin folds may contribute to cancer invasion.
Mesenchymal cancer cells such as MDA-MB-231 adhere to Coll I fibers at their front end and de-adhere at their rear end during migration. The adhesion-deadhesion dynamics is regulated through actin cytoskeletal dynamics which in turn is mediated through the phosphorylation of the Focal Adhesion Kinase (FAK) (35,36). The proinvasive functions of S100A4, proteomically identified to be under the regulation of Galectin-9 levels has been proposed to be associated with its localization and possible interaction with myosin (37) and cytoskeletal elements such as actin (38) and tubulin (39). Our demonstration that Galectin-9 driven accentuation of FAK signaling and its resultant potentiation of S100A4 gene expression, therefore suggests that its effect on the cytoskeletal dynamics of migrating breast epithelia maybe multilevel in nature. Moreover, S100A4 has also been shown to enhance FAK signaling in pancreatic cancer (40), which when taken together with our demonstration of the converse,

RNA isolation:
Total RNA was isolated using RNAiso Plus reagent from TaKaRa (TaKaRa, 9108) as mentioned in the product manual. Briefly, the spent medium was aspirated, and cells were

3D invasion assay:
3D invasion assay was performed as described in our earlier work (23 Ethanol, 1x 10min 80% Ethanol, 1x 10 min 70% Ethanol and finally in distilled water for 10 min. Antigen retrieval was performed using citrate buffer pH 6.0 in microwave for 30 min and allowed to cool down to room temperature. Sections were blocked using 5% BSA and 0.01% for 24 h. The same Puromycin pressure was maintained for two passages and then increased to 5 µg/mL for another two passages before being withdrawn. Stably transduced cells were confirmed for expression using western blot or quantitative real-time PCR.
Synthesis procedures, and data for the synthesis of 2 and 3:

Galectin-9 inhibition:
The  were set at 10 ppm and 0.5 D, respectively. The protease used to generate peptides, that is, enzyme specificity was set for trypsin/P (cleavage at the C terminus of "K/R": unless followed by "P") along with maximum missed cleavages value of two. Carbamidomethyl on cysteine as fixed modification and oxidation of methionine and N-terminal acetylation were considered as variable modifications for database search. Both peptide spectrum match and protein false discovery rate were set to 0.01 FDR. Statistical analysis was performed by using in-house R script. Abundance value for each run (including all biological replicates) were filtered and imputed by using normal distribution. Log2 transformed abundance values were normalized using Z-score. ANOVA and t-test was performed based on P-value (threshold P < 0.05) to identify the significant proteins.

a. Collective cell invasion
Collective cell invasion was analyzed using ImageJ (43). Image at 0 hour was selected, and cluster area was obtained after auto threshold. This is termed as initial area (A). Similarly, cluster area was obtained at the endpoint, and the area is termed the final area (B). Area of collective cell invasion is calculated using the formula: (B-A)/(A).
b. Single-cell invasion Image at endpoint was obtained, and the number of single cells dispersed in Coll I was counted manually using cell counter plugin in ImageJ.
c. Velocity and accumulated distance Velocity and accumulated distance were calculated using the manual tracking plugin. Briefly, single-cells coming out of the cluster was tracked manually for the last 12 hour of the timelapse videography. Tracks obtained from such single cells were analyzed using the Chemotaxis tool by ibidi, Germany.

Statistical analysis:
All experiments were performed in at least duplicates and repeated thrice independently. All data is represented as mean±SEM unless specified. For statistical analysis, unpaired student's t-test with Welch's correction or One-way ANOVA with Dunnett's multiple comparisons was    showing significantly increased S100A4 expression upon Galectin-9 overexpression. E) Graph showing significantly decreased S100A4 expression upon genetic knockdown of Galectin-9 expression. F) Graph showing significantly decreased S100A4 expression upon functional inhibition of Galectin-9 using 2 (10 µM) and 3 (10 µM) inhibitors.