Galectin-3, a Druggable Vulnerability for KRAS-Addicted Cancers

Oncogenic KRAS drives a diverse set of cellular mechanisms that support tumor progression, from activation of its effectors MEK and AKT that broadly support cell survival and proliferation, to cellular processes representing specific functional advantages, including macropinocytosis (1) and redox balance (2). However, targeting KRAS has yet to achieve clinical success. Not only does the frequency of KRAS mutation vary among cancer types [95% for pancreatic cancer vs. 25%–30% for non–small cell lung cancer (NSCLC)], but also individual KRAS-mutant tumors may develop KRAS indifference over the course of cancer progression (3, 4). Successfully targeting KRAS will require not only strategies to identify which tumors remain dependent on KRAS expression, but also new approaches to undercut the ability of mutant KRAS to trigger its diverse array of effectors that drive survival and tumor progression.

KRAS nanoclustering has emerged as a critical determinant of KRAS function (5). Unless oncogenic KRAS can be brought into proximity with its effectors at a relevant microdomain on the cell membrane, an activating KRAS mutation alone may not necessarily trigger specific downstream signaling events that generate KRAS addiction. Furthermore, KRAS signaling specificity is highly influenced by its association with distinct subsets of phospholipids that assemble at the plasma membrane (6), providing an explanation for the ability of KRAS to provoke diverse signaling outputs.

In matrix-adherent cells, KRAS is recruited into membrane nanoclusters by a variety of cell surface receptors that serve as interaction platforms to drive cell survival and proliferation. The redundancy in receptors capable of mediating KRAS clustering may explain why matrix-adherent cells can easily switch dependence between pathways when one is inhibited by a given targeted therapeutic. However, in the absence of matrix adhesion, epithelial cell surface receptors poorly cluster, as has been demonstrated for EGFR (7). Because the ability of a cell to exhibit anchorage-independent growth is a hallmark of tumor progression and metastasis (8, 9), we propose KRAS-containing signaling complexes that form in nonadherent cells as therapeutic targets to reverse anchorage independence, thereby containing the spread of malignant cells.

Integrin αvβ3 is unique among integrins for its ability to cluster on the surface of nonadherent tumor cells where it contributes to anchorage-independent growth (10). This may account for the association of αvβ3 expression with tumor progression, metastasis, and poor survival for a wide range of cancers (11). For example, expression of αvβ3 is increased from 10% in the primary tumor to 24% in metastases for lung cancer (12). In the unligated state, integrin αvβ3 is able to cluster and recruit tyrosine kinase signaling molecules to provide cancer cells with growth and survival signals across diverse microenvironments (13). Anchorage-independent clustering of integrin αvβ3 is influenced by galectin-3, a carbohydrate-binding protein that binds N-glycans on the integrin's extracellular domain (14), and we recently reported this αvβ3/galectin-3 interaction as an important modulator of EGFR inhibitor resistance by virtue of its ability to induce KRAS clustering in nonadherent cells (15).

We therefore asked whether integrin αvβ3/galectin-3 might represent an Achilles' heel for KRAS-mutant lung and pancreatic cancers, and if this could explain why only certain KRAS-mutant tumors remain addicted to KRAS for survival after tumor initiation. Not only do we demonstrate the potential of integrin αvβ3 as a biomarker to identify KRAS addiction, but we also provide a molecular explanation for this dependency.

Twenty-five percent to 30% of non–small cell lung cancers express mutant KRAS (16), yet only about half of these are addicted or depend on KRAS for their survival (17), thus representing an interesting model to investigate the molecular basis for KRAS addiction. Because we previously reported that the association between integrin αvβ3, galectin-3, and KRAS on the surface of lung cancer cells led to EGFR resistance (15), we considered whether this interaction might be a contributing factor to KRAS oncogene addiction. To do this, we selected three KRAS-mutant lung adenocarcinoma cell lines with detectable αvβ3 expression and three lacking αvβ3 expression to evaluate KRAS dependency on anchorage-independent growth (Fig. 1A; Supplementary Fig. S1A; Supplementary Table S1). Notably, expression of other integrin subunits (β1, β5, and αv) or KRAS-GTP activity was not linked to β3 expression. Strikingly, the αvβ3⁺ cell lines were addicted to both KRAS and galectin-3 (GAL3), whereas the αvβ3⁻ cell lines showed no such dependence (Fig. 1B; Supplementary Fig. S1B). Not only was integrin αvβ3 necessary for 3-D growth (Fig. 1C; Supplementary Fig. S1C and S1D), but its ectopic expression was sufficient to render cells dependent on both KRAS and galectin-3 (Fig. 1D; Supplementary Fig. S1C and S1E). Whereas 2-D anchorage-dependent growth of αvβ3-positive cells did not require expression of KRAS, galectin-3, or integrin β3, αvβ3-negative cells did require KRAS for 2-D growth (Fig. 1E). Together, these findings indicate that KRAS addiction arises from tumor cell reliance on an integrin αvβ3/galectin-3 pathway that supports anchorage-independent growth in 3-D, but which is dispensable in matrix-adherent cells.

We therefore considered the prevalence of the αvβ3-positive/KRAS-mutant subpopulation in human lung cancer. For 86 KRAS-mutant lung cancer patient-derived xenograft (PDX) models in the Crown Biosciences HuPrime database, 21% show significant integrin β3 (ITGB3) mRNA expression. Because the function of the integrin αvβ3 heterodimer requires the β3 subunit to couple with the αv subunit on the cell surface, we also screened 13 of these PDX models for tumor cell αvβ3 protein expression by immunostaining tumor sections (Supplementary Table S2). We found a wide range of αvβ3 expression on tumor cells, measured as the percent-positive cells per field (Fig. 1F). As for our panel of lung cancer cell lines, expression of αvβ3 was not linked to any particular KRAS mutation (Supplementary Tables S1 and S2). As expected, we also observed αvβ3 expression on tumor-associated stromal cells and blood vessels (11). For two PDX models with heterogeneous tumor cell expression of αvβ3 (PDX-8 and PDX-9), we confirmed that knockdown of either KRAS, galectin-3, or β3 significantly decreased anchorage-independent growth (Fig. 1G; Supplementary Fig. S1F). Using flow cytometry to sort the heterogeneous tumors for β3-hi and β3-lo populations, we show that only the β3-hi population is sensitive to knockdown of either KRAS or galectin-3 (Fig. 1G; Supplementary Fig. S1F). This work reveals an αvβ3-positive subset of KRAS-mutant lung cancer that shows KRAS addiction in the context of anchorage-independent growth.

Although KRAS signals through multiple effector pathways, we find that β3 expression is linked to the activation of phospho-AKT (pAKT), but not phospho-ERK (pERK; Fig. 1H). Although knockdown of KRAS decreases both pAKT and pERK, knockdown of either β3 or galectin-3 primarily suppresses the ability of mutant KRAS to activate AKT only (Fig. 1I). These findings are in accordance with recent studies that highlight “context-dependent” functions of KRAS (18).

To examine these relationships in the absence of other oncogenic drivers, we generated mouse embryonic fibroblasts (MEF) from animals expressing wild-type Kras, mutant Kras^G12D, galectin-3 (Lgals3) knockout, or the combination. Whereas KRAS activity driven by the G12D mutation did not require galectin-3 expression (Fig. 1J), the transforming potential of the Kras^G12D mutation requires galectin-3 as does phosphorylation of AKT (Fig. 1K; Supplementary Fig. S1H). Together, these findings suggest that αvβ3-positive cells are uniquely addicted to mutant KRAS for anchorage-independent growth, and point to galectin-3 as a critical mediator of this activity.

One recently described feature of KRAS-mutant cancer is an enhanced ability for macropinocytosis (1), a process that provides tumor cells with the ability to consume protein from their in vivo microenvironment as a unique source of amino acids (19, 20). We therefore asked if this functional benefit was common to all KRAS-mutant cells, or if αvβ3/galectin-3 might influence whether a given cell developed an addiction to this advantage. To do this, we measured cellular uptake of high molecular weight TMR-dextran and quantified the fluorescent signal emitted by degradation of a self-quenched albumin. As observed for the addiction to both KRAS and galectin-3, αvβ3⁺ cells showed markedly enhanced ethylisopropylamiloride (EIPA)–sensitive nutrient uptake compared with αvβ3⁻ cells growing in suspension (Fig. 2A and B) but not in 2-D adherent culture conditions (Supplementary Fig. S2A). We selected H727 and A549 cells to represent β3⁻ and β3⁺ groups, respectively, based on their intermediate levels of TMR-dextran uptake. Importantly, macropinocytosis was inhibited by knockdown of KRAS, β3, or galectin-3 in A549 cells, and enhanced by ectopic β3 expression in H727 cells (Fig. 2C; Supplementary Fig. S2B). For the PDX models with heterogeneous αvβ3 expression, individual cells with high αvβ3 expression showed the highest levels of macropinocytosis by flow cytometry analysis (Fig. 2D).

Consistent with these findings, treatment with EIPA, an inhibitor of macropinocytosis that does not affect other endocytic pathways (21), suppressed anchorage-independent growth of αvβ3⁺, but not αvβ3⁻, cells (Fig. 2E). In contrast, an inhibitor of clathrin-mediated endocytosis had an equivalent effect on all cells (Fig. 2E). As further evidence that αvβ3 is a critical determinant of this behavior, ectopic β3 expression was sufficient to enhance H727 cell sensitivity to EIPA (Fig. 2F), and the β3-positive population of PDX cells was significantly more sensitive to EIPA than the β3-negative population (Fig. 2G; Supplementary Fig. S2C). Together, these findings suggest that macropinocytosis does not contribute to the anchorage independence of KRAS-mutant lung cancer cells unless both integrin αvβ3 and galectin-3 are expressed.

Considering that the Kras^G12D mutation was required for anchorage-independent growth of MEFs with endogenous αvβ3 expression (Fig. 1J and K), we asked whether sensitivity to EIPA may also require Kras mutation. Whereas fibroblasts with wild-type Kras were largely insensitive to EIPA, expression of the Kras^G12D mutation was sufficient to induce an addiction to macropinocytosis that, as for anchorage-independent growth, required expression of galectin-3 (Fig. 2H). Together, these findings indicate that the enhanced capacity for anchorage-independent growth observed for αvβ3⁺ KRAS-mutant lung cancer cells can be attributed in part to their ability to use macropinocytosis for nutrient uptake. Because this process provides an additional energy route to support growth within nutrient-poor environments, blocking this pathway using EIPA or by targeting galectin-3 provides new options to target the unique vulnerabilities of αvβ3-expressing KRAS-mutant lung cancers.

In addition to macropinocytosis, oncogenic KRAS has recently been linked to oxidative stress detoxification during tumorigenesis (2). Because redox homeostasis has also been implicated in the adaptation to anchorage independence (22), we considered cellular detoxification as an additional hallmark vulnerability of KRAS-addicted lung cancer cells. Consistent with this notion, KRAS knockdown in αvβ3⁺ cells strongly induced accumulation of cellular reactive oxygen species (ROS), visualized by an increase in 8-oxo-dG or MitoSOX Red probes by immunofluorescence staining (Fig. 3A). In fact, only αvβ3⁺ cells could manage the oxidative stress induced by hydrogen peroxide during anchorage-independent growth (Fig. 3B), suggesting a potential link between stress tolerance and the combination of mutant KRAS and αvβ3 expression. Indeed, expressing β3 in H727 cells increased their tolerance to this form of oxidative stress (Fig. 3C). As further evidence, the β3-positive population of the PDX-9 tumor shows lower ROS levels than the unsorted or β3-negative cells (Fig. 3D), and integrin αvβ3 and galectin-3 were both necessary and sufficient to maintain low ROS levels in KRAS-mutant cells (Fig. 3E). These findings demonstrate that KRAS-addicted lung cancer cells require both galectin-3 and integrin αvβ3 to counteract the effects of oxidative stress, an advantage that may link KRAS addiction, tumor progression, and therapy resistance.

Our findings reveal that the impact of mutant KRAS in lung cancer cells is dependent on both integrin αvβ3 and galectin-3, and we demonstrated the functional implications of this for both macropinocytosis and redox balance. Considering that this pathway leads to various biological functions and may involve multiple effectors that could vary among individual tumors, we reasoned that disrupting signaling at the highest upstream point should provide the most robust effect. Although we previously reported a biochemical association among integrin αvβ3, KRAS, and galectin-3 (15), it was not clear whether there was direct binding among these factors, and how such an interaction might be prevented. Using a cell-free binding assay, we report here that galectin-3 directly binds to αvβ3 in dose-dependent and saturable manner, and this can be blocked by competitive binding with lactose, a sugar that mimics galactose binding to galectin-3 (Fig. 4A and B). Furthermore, we demonstrate that this αvβ3/galectin-3 interaction can be blocked with the anti–galectin-3 drug GCS-100 (Fig. 4C), a galectin-3–binding polysaccharide derived from citrus pectin that has shown a good safety profile and activity in early clinical phase trials for cancer (23–25). For αvβ3-expressing lung cancer cells growing in suspension, GCS-100 decreases the cell surface expression of galectin-3 (Fig. 4D). This indicates that soluble extracellular galectin-3 cannot bind to any of its cell surface receptors (including αvβ3) in the presence of the GCS-100 drug. Thus, this inhibitor provides a novel opportunity to perturb mutant KRAS/galectin-3/αvβ3 complex using an upstream approach.

Sensitivity to GCS-100 is dependent on integrin αvβ3, because cells positive for β3 expression were sensitive to the drug, whereas β3-negative cells were resistant (Fig. 4E). Furthermore, the β3-positive population of cells sorted from a heterogeneous PDX-9 lung tumor were more sensitive to the drug relative to the β3-negative or unsorted populations (Fig. 4F). We next asked whether the effects of GCS-100 were restricted to cells expressing mutant as opposed to wild-type Kras. Interestingly, Kras^G12D MEFs (which express endogenous αvβ3) are more sensitive to GCS-100 than MEFs isolated from Kras wild-type mice, and we confirmed that expression of galectin-3 is required for this effect (Fig. 4G). Treating established αvβ3-positive lung cancer PDX tumors with GCS-100 significantly decreased tumor growth (Fig. 5A), and this was associated with increased apoptosis evaluated by TUNEL staining (Fig. 5B).

We therefore considered whether GCS-100 could slow tumor progression for a spontaneous mouse model of KRAS-driven lung cancer. Three months after intratracheal adeno-Cre delivery to LSL-Kras^G12D mice, we confirmed the presence of lung tumors with integrin αvβ3 expression on tumor cells and pathologic evidence of regions containing adenoma, dysplasia, and adenocarcinoma (Fig. 5C). Not only did mice treated for 1 month with GCS-100 show a significantly smaller tumor burden, but this agent prevented tumor progression, as there was no evidence of adenocarcinoma in the mice (Fig. 5D and E). Importantly, treated tumors showed more apoptosis than controls (Fig. 5F). This finding demonstrates that galectin-3 inhibition has the potential to alter the phenotype of KRAS-driven tumors to have a marked impact on tumor progression in immune-competent mice. In concordance with our hypothesis that tumor cell addiction to mutant KRAS is modulated by αvβ3/galectin-3, the ability of GCS-100 to disrupt this interaction suggests its potential application as a therapeutic agent for KRAS-mutant cancers.

We therefore asked whether αvβ3 expression might predict GCS-100 sensitivity for pancreatic ductal adenocarcinoma, a cancer characterized by more than 95% frequency of KRAS mutation (26). As with lung cancer, pancreatic carcinoma cell lines were varied in their expression of integrin β3, and β3 expression was associated with phosphorylation of AKT for cells in 3-D, but not 2-D (Fig. 6A). GCS-100 shows αvβ3-dependent activity in vitro (Fig. 6B) and in vivo for pancreatic cancer xenografts (Fig. 6C) and PDX models (Fig. 6D) that are driven by oncogenic KRAS mutations. As for the lung cancer models, the antitumor activity of GCS-100 appears to involve an induction of cell death (TUNEL staining) without altering the extent of cell proliferation (Ki67 staining; Fig. 6E). Thus, in both immune-compromised and immune-competent mice, inhibiting galectin-3 with GCS-100 selectively inhibits the growth of KRAS-mutant tumors expressing integrin αvβ3.

Previous studies established a role for galectin-3 in the clustering of integrin αvβ3 on endothelial cells (14), and we reported that galectin-3 knockdown prevented the clustering of αvβ3, as well as the recruitment of KRAS, in cancer cells (15). Here, we demonstrate how the biochemical association between β3/KRAS can be disrupted pharmacologically by using the galectin-3 inhibitor GCS-100 (Fig. 7A and B). A consequence of this is a marked decrease in pAKT, whereas the drug does not affect pERK or KRAS-GTP activity (Fig. 7A and B). In fact, pAKT represents a very clear readout for activity of this drug in αvβ3-expressing tumors (Fig. 7C, D, and E). By preventing the ability of integrin αvβ3 to associate with KRAS, GCS-100 blocks the mutant KRAS-mediated survival advantages to which αvβ3-expressing cells have become addicted. Accordingly, treatment with GCS-100 not only prevents the cells from engulfing nutrients via macropinocytosis (Fig. 7F and G), but also increases the ROS levels for KRAS-mutant cells in 3-D culture (Fig. 7H) and tumor xenografts and PDX tumors in vivo (Fig. 7I).

Our findings indicate that lung adenocarcinoma addiction to oncogenic KRAS depends on galectin-3 to assemble an αvβ3/KRAS complex (Fig. 7J). Direct binding between αvβ3 and galectin-3 can convert cells expressing mutant KRAS to a KRAS-addicted state, and disrupting this complex with a galectin-3 inhibitor suppressed macropinocytosis, increased ROS, and selectively blocked the growth of KRAS-addicted lung and pancreatic cancer in mice. These findings provide a rationale for the design of clinical trials enabling the selective therapeutic targeting of KRAS-addicted cancers with the use of αvβ3 as a biomarker for these tumors.

Despite intense efforts to target KRAS or its key effectors, compensatory pathways have limited the ability of such agents to achieve clinical impact (26, 27). Although KRAS drives tumorigenesis of lung and pancreatic cancer, we and others (3) find that only certain KRAS-mutant lung cancers continue to need KRAS during tumor progression. Our findings provide a molecular explanation for KRAS addiction driven by αvβ3/galectin-3 and link this to the ability of mutant KRAS to initiate both macropinocytosis and ROS neutralization.

Alterations in nutrient uptake and metabolic reprogramming provide tumor cells with an enhanced ability to survive within an adverse environment. In particular, active RAS has been linked to a hypermacropinocytotic phenotype that allows cells to engulf nutrients and alter their metabolic dependencies (1, 28). In contrast, we find that only a subset of KRAS-mutant lung cancer cells show enhanced macropinocytosis, and that expression of αvβ3 defines this subset. In fact, cells that rely on KRAS-mediated macropinocytosis also depend on the ability of αvβ3 to cluster on the surface of cells with galectin-3, forming a complex that recruits active KRAS to the membrane. This is supported by the finding that only tumors with mutant KRAS and αvβ3 expression are sensitive to the macropinocytosis inhibitor EIPA.

In addition to driving macropinocytosis dependency, the αvβ3/galectin-3 complex is required to maintain low mitochondrial ROS levels in KRAS-addicted cells. It is well appreciated that a tumor cell with enhanced cellular detoxification abilities, including some that are driven by active RAS (2), can survive and proliferate despite diverse environmental stresses or exposure to cancer therapy. It is therefore interesting to consider how individual cancer cells may be able to maintain redox balance, for example by upregulating antioxidant expression.

Notably, we propose integrin αvβ3 as a biomarker capable of identifying KRAS addiction (and GCS-100 sensitivity) independent from any other genetic alterations. Although the β3 subunit associates only with αv (ubiquitously expressed) and αIIb (predominantly in platelets; ref. 11), αv combines with multiple other β subunits to form heterodimers with diverse functions. As such, β3 expression is rate limiting for αvβ3 expression on virtually all epithelial cells. Differences in protein or mRNA expression of the β3 subunit are therefore sufficient to identify tumor cell addiction to this pathway, whereas measuring αv expression alone would not predict KRAS addiction. Similarly, galectin-3 protein expression is detectable in all cell lines and PDX models examined, suggesting that galectin-3 levels are not likely to predict sensitivity to a galectin-3 inhibitor. Within the tumor microenvironment, soluble galectin-3 released from stromal cells might interact with αvβ3 on the surface of a tumor cell. Taken together, our results suggest that the ability of galectin-3 to directly bind the β3 integrin ectodomain creates addiction to both galectin-3 (as a mediator of αvβ3 clustering in anchorage-independent cells) and KRAS (as a critical downstream effector that drives multiple survival advantages).

It is also interesting to point out that, among the αvβ3-positive KRAS-mutant cell lines we have analyzed in our study, A549 cells have mutant STK11 with wild-type TP53, whereas the H1792 cells express wild-type STK11 and mutant TP53. Despite these differences, both cell lines are addicted to KRAS and require integrin αvβ3/galectin-3. As such, targeting galectin-3 as a requisite for KRAS clustering in cells with αvβ3 expression may provide a more universal approach to disable the downstream effects that could vary between tumors with distinct genetic drivers of cancer.

Notably, our determination of KRAS dependency in 3-D culture varies from the designations previously reported for some of the same cell lines tested under 2-D culture (3, 29). We believe the key difference is that αvβ3 is the only integrin capable of clustering on the surface of nonadherent cells (11), whereas multiple receptors are able to recruit KRAS into focal-adhesion type signaling nodes that are supported by cell-matrix adhesion. As a result, we propose a unique role for mutant KRAS in promoting the survival of the αvβ3-positive subset of lung cancer cells in the setting of anchorage-independent growth, in vitro conditions that partially recapitulate the environmental stresses tumor cells must overcome during metastasis and tumor progression in vivo (10, 30). In fact, we show that ectopic expression of β3 is sufficient to drive KRAS addiction, suggesting αvβ3-negative cells are simply unable to engage this survival mechanism (i.e., KRAS cannot otherwise be clustered) in cells growing under anchorage-independent conditions. Our findings suggest that cooperation between integrin αvβ3 and galectin-3 provides an important contextual signal that is required for the function of mutant KRAS, and that disruption of this pathway could represent a unique vulnerability that could be targeted therapeutically.

In the context of 2-D growth, Singh and colleagues classify the β3-positive A549, SKLU1, and PANC1 cells as KRAS-independent and the β3-negative H727 and H441 cells as KRAS-dependent (3), and they show that A549/SKLU1/PANC1 cells are able to activate AKT in the absence of KRAS during 2-D growth, whereas we suggest that integrin αvβ3 is the only integrin capable of recruiting and associating with KRAS to activate AKT in 3-D. Although αvβ3 might drive KRAS independence in 2-D by activating AKT in a manner that circumvents KRAS, we focus primarily on 3-D growth conditions that mimic the in vivo tumor growth environment required for invasion and metastasis (31). As such, we believe that the growth advantage imparted by αvβ3/KRAS in 3-D can be explained by their ability to promote AKT activity that drives both macropinocytosis and redox balance, features that may be less critical for cells growing in a 2-D monolayer with media containing full serum.

Previous studies have shown that galectin-3, a metastasis-inducing protein expressed by tumor cells and inflammatory cells (32–34), not only binds in a multimeric manner to the oligosaccharides on the extracellular domain of integrin αvβ3 (14), but also associates directly or indirectly with KRAS (35), mediates KRAS clustering at the plasma membrane (36–38), and can augment KRAS activity (39, 40). In fact, we previously reported that αvβ3 and galectin-3 form a membrane complex with KRAS to promote tumor stemness and drug resistance (15). It is therefore remarkable that only certain KRAS-mutant lung cancer cells are inherently “addicted” to galectin-3 during anchorage-independent growth, a feature we attribute to its impact on αvβ3 recruitment of KRAS.

One of the most interesting outcomes of our study is that a clinically active galectin-3 inhibitor, GCS-100, has the capacity to undermine the survival of tumor cells that are αvβ3-positive and KRAS mutant in vitro and in vivo, including a spontaneous Kras-driven lung cancer model. In this model, GSC-100 actually prevented tumor progression as measured by dysplasia and adenoma but a complete absence of adenocarcinoma. In fact, among KRAS-mutant lung and pancreatic tumors, αvβ3 expression was necessary and sufficient for the effects of this drug. This activity was not only related to epithelial cancers, as galectin-3 was required to drive tumor formation among MEFs expressing mutant but not wild-type Kras. This suggests that oncogenic activation of KRAS is not sufficient to drive KRAS addiction, but that the recruitment of KRAS by αvβ3/galectin-3 is critically important.

The selective activation of AKT in αvβ3-expressing KRAS-mutant carcinoma cells may also contribute to KRAS addiction, considering that PI3K/AKT drives survival signaling. Although AKT has many effectors, its phosphorylation in cells and tumors treated with GCS-100 is markedly suppressed, suggesting that αvβ3/KRAS/galectin-3 is a driver of pAKT during anchorage-independent growth and tumor progression of KRAS-addicted tumors. Although pAKT provides a readout for activity of the galectin-3 inhibitor GCS-100, our work does not exclude the possibility that additional KRAS effectors may also contribute to the survival advantages driven by KRAS in αvβ3-expressing cells. As such, targeting KRAS function by virtue of its association with integrin αvβ3 represents a strategy to block multiple KRAS-driven functions including but not limited to macropinocytosis and redox balance.

Galectin-3 can modulate the function of multiple cell types important during for progression (33, 41), and the early development of galectin-3 inhibitors for cancer therapy focused on blocking the function of tumor-associated endothelial cells (42). Although we cannot rule out that the tumor growth inhibition observed for GCS-100 is exclusively due to its disruption of αvβ3 clustering on tumor cells, we do show a similar level of tumor growth inhibition in immune-compromised versus immune-competent mice (Fig. 5C–F), and we show selective drug activity for tumors with ectopic β3 expression but not the respective parental line (Fig. 6C). Together, these findings suggest that blockade of the αvβ3/galectin-3 interaction on the surface of tumor cells represents a novel approach to target KRAS-addicted cancers.

Although mutant KRAS is involved both in lung cancer tumorigenesis and progression, it becomes dispensable for the progression of a subset of tumors. Here, we identify lung cancers that remain addicted to KRAS by virtue of their dependence on key KRAS-mediated survival advantages, including nutrient uptake and redox balance. Although previous studies link macropinocytosis to oncogenic KRAS in pancreas cancer (1), we find that only certain KRAS-mutant lung cancers (those expressing both galectin-3 and integrin αvβ3) rely on this process for anchorage-independent growth. Although KRAS is implicated as a driver of redox gene expression (2, 43), we show that αvβ3-negative cells are more sensitive to oxidative stress and show higher basal ROS levels than αvβ3-positive cancer cells. In this respect, our findings provide a new mechanistic explanation to account for KRAS addiction by linking established KRAS functions to the ability of KRAS to form a complex with integrin αvβ3 in anchorage-independent cells. Because tumors addicted to this pathway for survival are highly sensitive to galectin-3 inhibition, this strategy represents an approach to treat the subset of KRAS-mutant lung cancers for which there are currently few viable options (Fig. 7J). Interrupting this pathway at the level of αvβ3/KRAS association may prove more promising than strategies to block individual KRAS-driven functions, which are likely not limited to only macropinocytosis and redox balance.

Cells were confirmed Mycoplasma-free before experiments. Cells were purchased in 2013 (A549, H441, and PANC-1), 2014 (SKLU-1 and H1792), and 2015 (A427 and H727), from the American Type Culture Collection and grown in recommended medium supplemented with 10% FBS and glutamine (RPMI for A549, H441, H1792, and H727; EMEM for SKLU-1 and A427; DMEM for PANC-1). FG and FGβ3 cells (DMEM) were obtained as described (10). Identity of the A549 cell line was confirmed using DNA fingerprinting. In all experiments, each cell line was passaged less than 15 times. Unless stated otherwise, tissue culture plates were coated with a 6% poly-HEMA solution to create a nonadherent condition wherein cells tend to grow in floating clumps rather than attach to the bottom of the wells.

EIPA, hydrogen peroxide, chlorpromazine, lactose, and sucrose were purchased from Sigma-Aldrich. The galectin-3 inhibitor GCS-100 was provided by La Jolla Pharmaceutical Company.

Cells were transfected with vector control or integrin β3 using a lentiviral system as described (15). For knockdown experiments, cells were transfected with shRNA (Open Biosystems) using a lentiviral system and the constructs listed in Supplementary Table S3. Lentiviruses were produced by cotransfection of 293T cells with lentiviral backbone constructs and packaging vectors (ps-PAX2 and VSVG) using Lipofectamine 3000 (Thermo Fisher). Supernatant was collected 48 and 72 hours after transfection and resuspended in an appropriate volume of OptiMEM (Gibco). Knockdowns were confirmed by immunoblot.

Immunoblotting was performed as described (15) for lysates generated using RIPA, Triton, or NP-40 buffers. Protein (25 μg) was boiled in Laemmli buffer and resolved on an 8% to 15% gel. Primary antibodies used were as follows: integrin β3 [Cell Signaling Technology (CST); 13166S; 1:1,000], KRAS (Abgent; AT2650a; 1:1,000), galectin-3 (Santa Cruz Biotechnology; sc-20157; 1:500), pAKT (CST; 2606S; 1:1,000), AKT (CST; 2938S; 1:1,000), pERK (CST; 9101S; 1:2,000), ERK1/2 (CST; 4695S; 1:1,000), HSP70 (Santa Cruz Biotechnology; sc-221731; 1:1,000), GAPDH (CST; 2118S; 1:5,000), and HSP90 (Santa Cruz Biotechnology; sc-7947; 1:1,000).

Levels of activated RAS in cell lysates were determined using a RAS Activation Assay Kit (EMD Millipore; # 17-218) following the manufacturer's instructions with anti-KRAS mouse monoclonal antibody (Abgent; AT2650a).

Lysates from cells grown in 3-D were generated using MLB buffer (Millipore; # 20-168). Immunoprecipitation experiments were performed using 500 μg of protein and anti-KRAS (Abgent; AT2650a).

Soft-agar assays were performed as described (10) for 14 to 21 days with weekly media replacement. Cells were grown on poly-HEMA–coated plates for 3 to 6 days in serum-free media, and CellTiter-Glo viability assays (Promega) were performed according to the manufacturer's instructions (44). To evaluate growth in a clonogenic assay, MEFs were seeded in a 96-well plate in complete media. After 1 day, serum-free media containing test agents were added and refreshed twice per week, and crystal violet staining was performed after 2 weeks.

MitoSOX Red (Thermo Fisher) was used according to the manufacturer's instructions to assess mitochondrial oxidation by superoxide in live cells using confocal microscopy (45) or flow cytometry. Alternatively, cells were immunostained with mouse monoclonal 8 hydroxy 2′ deoxyguanosine (Abcam #ab48508) as a measure of oxidative damage to DNA. Images were acquired using confocal or light microscopy.

As described (1, 46), cells were grown in 3-D using poly-HEMA–coated plates in serum-free media for 3 hours. Macropinosomes were marked using a high-molecular TMR-dextran and/or DQ-Green BSA (Life Technologies) at a final concentration of 1 mg/mL for 1 hour at 37°C. Cells were rinsed in cold PBS, fixed in 3.7% formaldehyde, and coverslips mounted using Hard Set (Vector Labs). Images were captured using a Nikon Eclipse C1 confocal microscope with 1.4 NA 60x oil-immersion lens and minimum pinhole (30 μm), and analyzed using the “Analyze Particles” feature in ImageJ. Particle area per cell was determined from at least three randomly selected fields, each containing approximately 75 to 150 cells depending on cell size and seeding density. As described for the ex vivo assay (47), tumors were cut into 3-mm cubes, immersed in serum-free RPMI with 1 mg/mL of FITC-dextran at 37°C for 1 hour, rinsed in PBS, and frozen in optimal cutting temperature compound.

PDX tumors were harvested from mice, washed with cold PBS containing antibiotics, chopped with a sterile blade, and incubated in 0.001% DNase (Sigma-Aldrich), 1 mg/mL collagenase/dispase (Roche), 200 U/mL penicillin, and 200 mg/mL streptomycin in DMEM/F12 medium in a 37°C water bath for 1 hour with intermittent shaking. The suspensions were repeatedly triturated, passed through 70- and 40-mm cell strainers, and centrifuged at 300 g for 5 minutes at 4°C. Cells were resuspended in red blood cell lysis buffer for 4 minutes at room temperature with intermittent shaking before resuspension in serum-free medium. Viability was evaluated by exclusion of Trypan blue dye (Thermo Fisher). For some experiments, cells were sorted into β3⁺ and β3⁻ populations using flow cytometry as described (15).

Single-cell suspensions from PDX tumors were processed as described (15) using anti-αvβ3 (LM609; Millipore; 1:1,000) followed by AlexaFluor-labeled secondary antibodies (Invitrogen) with DAPI (Thermo Fisher) as a nuclear marker. Samples were imaged on a Nikon Eclipse C1 confocal microscope with 1.4 NA 60x oil-immersion lens, using minimum pinhole (30 μm). For FACS analysis, cells were stained with galectin-3 (BioLegend; 125401; 0.5 mg/mL diluted 1:200), LM609 (1:1,000), or isotype control.

Tumor sections were subjected to hematoxylin and eosin (H&E) staining following standard protocols. TUNEL staining for apoptosis was performed using the APO-BrdU-IHC assay (Phoenix Flow Systems; #AH1001). Sections (5 μm thick) of paraffin-embedded tumors were immunostained using the VECTASTAIN Elite ABC HRP Kit (Vector Labs). Primary antibodies for human tumors included integrin β3 (CST; #13166), pAKT (CST; # 3787), and 8-oxo-d-Guo (Abcam; #ab48508). Mouse tumors were stained using integrin β3 (Abcam; #ab119992). H&E-stained lung tissues were imaged on a NanoZoomer Slide Scanning System (Hamamatsu). Tumor burden was measured with the ImageJ Threshold Color plugin.

Ninety-six–well plates were coated with purified human integrin αvβ3 (Millipore CC1021; 0.5 μg in 100 μL), incubated at 4°C overnight, and blocked with 50 mg/mL BSA for 90 minutes at 30°C. Recombinant human galectin-3 (R&D Systems 1154-GA-050; 1 μg/well) and test agents were added for a total volume of 50 μL, then incubated for 4 hours at 30°C. Wells were washed, fixed with 2% PFA in PBS for 15 minutes at room temperature, washed, and incubated with rat monoclonal galectin-3 antibody (BioLegend 125401; 0.5 mg/mL diluted 1:100) for 30 minutes on ice. After washing, wells were incubated with secondary antibody (Life Technologies; A21210, AF488 rabbit anti-rat IgG diluted 1:200) for 30 minutes on ice. After washing, fluorescence was read using a Tecan Infinite M200 (excitation 485, emission 538) to quantify αvβ3/galectin-3 binding.

All research was conducted under protocol S05018 and approved by the UCSD Institutional Animal Care and Use Committee. All studies are in accordance with the NIH Guide for the Care and Use of Laboratory Animals.

As previously published for LSL-^KrasG12D mice, intratracheal delivery of adeno-Cre induces oncogenic Kras in lung airway cells, leading to multifocal adenocarcinomas and a median survival of about 6 months (48). Starting with tumors established for 3 months in adult B6.129 LSL-Kras^G12D (Jackson Labs; 008179) mice of either gender, systemic dosing with vehicle or 20 mg/kg GCS-100 3 times a week by i.p. injection was performed for 1 additional month. Tumors were harvested, fixed, and stained for tumor burden analysis using H&E. Scanned slides were scored by a pathologist for the presence of adenomas with benign proliferation versus preneoplastic lesions (dysplasia) and adenocarcinomas, as defined (49, 50).

Subcutaneous PDX tumors were grown in 6- to 8-week-old female SCID mice as reported (15). Molecular Response LLC (now Crown Bioscience) provided slides containing tumor sections of the PDX models listed in Supplementary Table S2 to allow immunohistologic screening for integrin αvβ3 expression, as well as cryovials of the 2001030397 PDX model for use in vitro and implantation in vivo. The UC San Diego Moores Cancer Center Biorepository and Jackson Labs provided the MCCT-009.4-LG1202F PDX model. The pancreatic PDX model was provided by Dr. Andrew Lowy at UCSD under UCSD Institutional Review Board protocol #090401. FG and FGβ3 human pancreatic carcinoma cells (5 × 10⁶ tumor cells in 200 μL PBS) were injected subcutaneously to the flank of 6- to 8-week-old female nu/nu mice. Animals bearing 150 to 250 mm³ tumors were randomly allocated for treatment with vehicle or 20 mg/kg GCS-100 i.p. 3 times per week.

All analyses were carried out using Prism software (GraphPad). One-way ANOVA or Welch t tests were used to calculate significance using P < 0.05. More than two independent experiments were carried out for all in vitro tests. All of the figures showing immunoblots or micrographs were independently repeated 3 times. The number of independent experiments versus technical replicates is specified in each figure legend.

L. Seguin reports receiving commercial research support from La Jolla Pharmaceuticals. D.A. Cheresh reports receiving commercial research support from La Jolla Pharmaceuticals and is a consultant/advisory board member for the same. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Seguin, J.S. Desgrosellier, S.M. Weis, D.A. Cheresh

Development of methodology: L. Seguin, M.F. Camargo, E. Cosset

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Seguin, M.F. Camargo, H.I. Wettersten, S. Kato, J.S. Desgrosellier, T. von Schalscha, K.C. Elliott

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Seguin, H.I. Wettersten, S. Kato, T. von Schalscha, S.M. Weis, D.A. Cheresh

Writing, review, and/or revision of the manuscript: L. Seguin, T. von Schalscha, S.M. Weis, D.A. Cheresh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Seguin, T. von Schalscha, K.C. Elliott, J. Lesperance

Study supervision: S.M. Weis, D.A. Cheresh

We thank Maricel Gozo, Isabelle Tancioni, David Tarin, Chloe Feral, and Alexandre Larange for helpful discussions. We also thank Molecular Response, Crown Bioscience, and the UC San Diego Moores Cancer Center Biorepository for providing PDX models.

D.A. Cheresh received grant support for this project from the NIH/NCI (R01CA45726), the California Institute for Regenerative Medicine (RB5-06978), and a Translational Research Grant from The V Foundation for Cancer Research. L. Seguin received a postdoctoral fellowship from Fondation ARC pour la Recherche sur le Cancer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.