Inhibiting SUMO1-mediated SUMOylation induces autophagy-mediated cancer cell death and reduces tumour cell invasion via RAC1

The post-translational addition of the small ubiquitin-related modifier (SUMO) peptide is now established as one of the key regulatory modifications in eukaryotic cells. SUMOylation involves the reversible binding of a SUMO peptide to a lysine residue in the target protein and, to date, four different SUMO isoforms have been identified: SUMO1, SUMO2 and SUMO3 (denoted SUMO2/3 because they have a high degree of similarity), and SUMO4 (Saitoh and Hinchey, 2000). The addition of these peptides is mediated by an enzyme cascade that includes an activating enzyme (heterodimer SAE1/2), an E2-conjugating enzyme (UBC9) and an E3 ligase (RANBP2 from the SIZ/PIAS family and members of the ZNF451 family) (Hay, 2005; Pichler et al., 2017), and the substrates can be modified by adding a single SUMO moiety, multiple SUMOs or SUMO chains (Pichler et al., 2017). In addition, specific SUMO-modified proteins can be deSUMOylated by a group of sentrin/SUMO-specific proteases (SENPs) (Hay, 2007).

The conjugation of a SUMO moiety affects proteins distinctly, modifying their activity, subcellular localization or stability. Many cellular pathways are influenced by reversible SUMOylation, such as those that influence chromatin organization, transcription, DNA repair, macromolecular assembly, protein homeostasis, trafficking and signal transduction (Geiss-Friedlander and Melchior, 2007). Failure to SUMOylate specific targets and the global misregulation of SUMOylation has been linked to different diseases, including cancer and heart failure (Flotho and Melchior, 2013). Therefore, interfering with the SUMOylation machinery could represent a novel therapeutic approach in the management of some diseases.

Ginkgolic acids (GAs) are a group of alkyl phenols found in crude extracts of Ginkgo biloba leaves, an ancient gymnosperm species now distributed globally (Mahadevan and Park, 2008). There are several molecular species of GA; these have a different length for their alkyl group within the main structure of the molecule (C13:0, C15:1 and C17:1). GAs display anti-cancer activity, and in several studies GA has been shown to inhibit the growth and invasion of a number of cancer cell types, including pancreatic, liver, pharyngeal and colon cancer (Qiao et al., 2017). While the mode of action of these compounds is still poorly understood, GA C15:1 has been shown to directly bind to E1 activating enzymes and impair the formation of the E1–SUMO1 intermediate (Fukuda et al., 2009). However, it remains to be clarified whether the anti-cancer activity of GAs depends on inhibition of the SUMO machinery or if additional mechanisms are involved in this effect.

RAC1 is a member of the Rho family of small GTPases that act as molecular switches to control a wide array of cellular events. RAC1 activity can modulate the cytoskeleton, which is critical for a number of cellular activities such as phagocytosis, mesenchymal-like migration, axon growth, adhesion, cell differentiation and cell death mediated by reactive oxygen species (ROS) (Acevedo and Gonzalez-Billault, 2018). RAC1 also plays an important role in moderating other signalling pathways that influence cell growth and the cell cycle (Mettouchi et al., 2001; Olson et al., 1995), the formation of cell–cell adhesions (Daugaard et al., 2013) and contact inhibition (Nobes and Hall, 1995). These RAC1-mediated activities appear to be central to the processes that underlie malignant transformation, including tumorigenesis, angiogenesis, invasion and metastasis (Mack et al., 2011).

The RAC1 GTPase binds to either GTP or GDP, the exchange of which controls its activation. RAC1 is inactive in the GDP-bound state and it is activated upon exchange of its GDP for GTP, enabling downstream signalling to proceed. RAC1 activity can be regulated through its association with several guanine nucleotide-exchange factors (GEFs) and GTPase-activating proteins (GAPs), these controlling the cycling between the GDP- and GTP-bound states. Furthermore, post-translational modifications (PTMs) of RAC1 can also regulate its activity. As such, modification of the C-terminal CAAX motif in RAC1 through the addition of either farnesyl or geranylgeranyl isoprenoid lipids increases its hydrophobicity, facilitating both its membrane localization and activation (Mack et al., 2011). Ubiquitin-like (UBL) modifications of RAC1 have also been shown to regulate its activity, including ubiquitylation (Castillo-Lluva et al., 2013) and SUMOylation (Castillo-Lluva et al., 2010), adding further complexity to the regulation of RAC1 signalling.

We observed RAC1 GTPase SUMOylation (RAC1-SUMO1) when the epithelial to mesenchymal transition (EMT) was induced by hepatocyte growth factor (HGF) (Castillo-Lluva et al., 2010). EMT involves changes in gene expression, and it is associated with a loss of cell polarity and an increase in cell invasiveness (Brabletz et al., 2018). The RAC1 GTPase plays an important role in the EMT programme (Ungefroren et al., 2018) and significantly, RAC1 SUMOylation is necessary for optimal cell migration when non-tumorigenic cells undergo EMT. Similarly, cancer cells also induce the EMT programme when they metastasize and invade other tissues (Brabletz et al., 2018), such that RAC1 SUMOylation could also play an important role in this context.

Here, we demonstrate that blockade of the SUMO1 conjugation pathway inhibits two of the cellular programmes that are activated during tumorigenesis, cancer cell survival and invasiveness. These effects are due to the activation of two independent mechanisms: the induction of autophagy-mediated cancer cell death through enhanced TRIB3 expression, and inhibition of RAC1-dependent cancer cell migration and invasion. Tumour cell invasion and metastasis are thought to be responsible for 90% of cancer-associated deaths. Thus, inhibiting SUMOylation could represent a novel therapeutic strategy to convert cancer from a mortal into a chronic disease.

As a first approach to investigate the effect of inhibiting the SUMO pathway on the tumorigenic properties of cancer cells, we analysed the effects of the natural compound GA C15:1 (hereafter referred to as GA), which blocks the SUMO pathway by inhibiting the formation of the E1–SUMO1 intermediate (Fukuda et al., 2009). Exposure of luminal (MCF7), triple-negative (MDA-MB-231) and HER2+ (BT474) breast cancer cells to a relatively low dose of GA (10 µM) reduced the overall protein SUMOylation in these cells by ∼40–50% (Fig. 1A). Moreover, this reduction in SUMO conjugation was more pronounced when these cells were exposed to a higher dose of GA (20 µM, Fig. S1A). When we then tested the effect of this compound on the survival of the different breast and prostate cancer cell lines, exposure to GA clearly diminished their viability in a dose-dependent manner (Fig. 1B; Fig. S1B). Hence, the effect of GA on cell viability appears to be common to different types of cancer cells. We identified the IC₅₀ of GA as 10 µM in MDA-MB-231 cells and 20 µM in MCF7 cells (the concentration that reduced the viability of cells by 50% compared with the control after a 2 day treatment; Fig. 1C; Fig. S1C). Accordingly, to study the mechanisms underlying these effects, we used sub-maximal doses of GA (10 µM) to enhance specificity and to avoid any potential toxicity associated with the excessive loss of protein SUMOylation. When we assessed the proliferation of cancer cells through the expression of the Ki67 marker, exposure to GA (10 µM) provoked a decrease in proliferation (Fig. 1D,E) and it reduced the levels of JUN (Fig. S1D), a transcription factor associated with the proliferation of breast cancer cells (Vleugel et al., 2006).

To confirm that the effect of GA observed was due to the inhibition of SUMOylation, we used a different SUMO1 inhibitor (2-D08) that blocks the transfer of SUMO from the E2 enzyme (UBC9) thioester conjugate to the substrate (Choi et al., 2018; Kim et al., 2014). Exposure of the triple-negative cell line (MDA-MB-231) to a relatively low dose of GA and 2-D08 reduced overall protein SUMOylation (Fig. 1F). While GA reduced the intensity of SUMO-conjugated proteins at 24 h, 2-D08 achieved a similar effect after a 48 h exposure. Accordingly, no differences in cell proliferation were observed in the presence of 2-D08 after 24 h, although cell viability was compromised after 48 h in the presence of this SUMO1 inhibitor (Fig. 1G,H).

To further confirm that SUMO1-mediated SUMOylation was responsible for the loss of cell viability observed in the presence of GA, we used a specific pool of siRNAs to deplete SUMO1 over 72 h (Fig. 1I) in the triple-negative MDA-MB-231 breast cancer cell line, a particularly aggressive subtype of breast cancer. As a result, we found that the depletion of SUMO1 reduced cell viability (Fig. 1J) and impaired cell proliferation (Fig. 1K,L).

Together, these results support the notion that inhibition of protein SUMOylation by using genetic or pharmacological approaches reduces the viability and the proliferation of cancer cells.

To understand how the SUMO pathway might affect the viability of cancer cells, we analysed the mechanisms activated in response to GA in the triple-negative MDA-MB-231 breast cancer cell line. We measured the rate of MDA-MB-231 apoptosis after a 48 h exposure to GA (10 µM), a time point at which cell viability was compromised (Fig. 1B). By using an Annexin V-FITC and propidium iodide (PI) staining assay, GA was seen to increase the proportion of apoptotic cells after 48 h of treatment (Fig. 2A,B) and indeed, an increase in cleaved caspase-3 and cleaved PARP was also evident in Western blots and by immunofluorescence after a 48 h exposure to GA (Fig. 2C–G). Similar results were also obtained with the SUMO inhibitor 2-D08 (Fig. S2A). Furthermore, we identified an increase in the apoptotic protein BAX at early time points (12 and 24 h, Fig. 2H), suggesting that early activation of the intrinsic mitochondrial pathway might be involved in the stimulation of apoptosis by GA.

We then investigated the mechanisms that might trigger this apoptosis at earlier time points, as other compounds have already been seen to promote apoptotic cancer cell death by stimulating autophagy (Fulda and Kögel, 2015; Salazar et al., 2009b). Autophagy is an essential homeostatic process that facilitates the transport of cellular components to the lysosomal degradation pathway. Depending on the cell context, stimulating autophagy can produce protective or cytotoxic effects (Wilde et al., 2018). We assessed whether the stimulation of autophagy might reduce tumour cell viability in a manner similar to that observed upon SUMO inhibition with GA. A 24 h exposure to GA triggered the accumulation of LC3II (the lipidated and autophagosome-associated form of LC3 family proteins), as assessed by western blotting, in both breast and prostate cancer cells, and as also evident in immunofluorescence analyses (Fig. 2I–L; Fig. S2B–D). At this time point, no change in cell viability had yet been observed. Similar results were also obtained with the SUMO inhibitor 2-D08 (Fig. S2E–G).

It is known that LC3 lipidation (to give LC3II) and its accumulation can be due to both the induction and blockade of autophagy (Klionsky et al., 2016). Pharmacological inhibition of the final step of autophagosome degradation (using E64d and pepstatin-A) led to the accumulation of more LC3II, indicating that GA activated autophagy rather than blocking it (Fig. 2M,N). Furthermore, shorter exposures to GA or 2-D08 induced the expression of regulators of autophagy, such as BECN1, ATG5, ATG1, ATG7 or AMBRA1 (Fig. 2O; Fig. S2H,I). Importantly, there was no decrease in cell viability when MDA-MB-231 cells were treated with GA, E64d and pepstatin-A (Fig. 2P), and the increase in apoptosis evoked by GA was also prevented (Fig. 2Q). Moreover, genetic depletion of three different key autophagy regulatory genes (ATG5, BECN1 and AMBRA1) appeared to rescue the apoptosis of these cells upon exposure to GA or 2-D08 (Fig. 2R; Fig. S2J–L).

Together, these observations support the idea that inhibiting SUMOylation induces autophagy-dependent cell death in breast cancer cell lines.

Several anti-cancer agents that stimulate autophagy-dependent cell death produce their effects by upregulating Tribbles pseudokinase-3 (TRIB3), including natural compounds of lipid origin that exhibit structural similarities to GA (Erazo et al., 2016; Salazar et al., 2009a; Vara et al., 2011). TRIB3 is an upstream regulator of the mammalian target of rapamycin complex 1 (mTORC1) (Salazar et al., 2015), the inhibition of which is considered to be the main mechanism triggering autophagy in many cell settings (Mizushima et al., 2008). Thus, we asked whether the stimulation of autophagy-mediated cell death through pharmacological blockade of the SUMO pathway might be related to changes in the expression of this protein. Exposure to GA increased the levels of TRIB3 in breast and prostate cancer cell lines (Fig. 3A,B; Fig. S3A), as did exposure to the 2-D08 SUMO inhibitor (Fig. 3C,D). Therefore, we hypothesized that the upregulation of TRIB3 might be the primary event leading to autophagy-mediated cell death in response to pharmacological inhibition of the SUMO pathway. In fact, stable knockdown of TRIB3 expression in MDA-MB-231 cells rendered these cells more resistant to GA-induced autophagy and cell death (Fig. 3E,F), preventing the accumulation of the cleaved forms of caspase-3 and PARP (two well established read-outs of apoptosis) (Fig. 3G,H).

Hence, these observations led us to conclude that TRIB3 induction is involved in the stimulation of autophagy-mediated apoptotic cell death upon pharmacological inhibition of the SUMO pathway.

To verify that the above-described effects were based on the inhibition of SUMO1, we knocked down SUMO1 expression in MDA-MB-231 cells by using a pool of 3 SUMO1-selective siRNAs. After a 72-h treatment, we found that the depletion of SUMO1 increased the accumulation of LC3 dots (reflecting the presence of autophagosomes in the cell), as well as the LC3II accumulated in these cells (Fig. 4A–D). Furthermore, we found that SUMO1 depletion induced a clear TRIB3 upregulation in MDA-MB-231 cells (Fig. 4E,F). Similar results were obtained when we used a different pool of two SUMO1-selective siRNAs (Fig. S3B) or when we depleted the expression of the E2–SUMO ligase, UBC9 (Fig. 4J). Moreover, we observed an increase in cell death and apoptosis after a 5 day depletion of SUMO1 that was prevented when the last step of autophagic degradation was pharmacologically blocked with E64d and pepstatin-A (Fig. 4G–I). By contrast, no differences were observed after SUMO2/3 depletion on TRIB3 expression, autophagy or apoptosis (Fig. 4K–M).

Taken together, these findings support the idea that pharmacological or genetic inhibition of the SUMO1 pathway activates autophagy-mediated cancer cell death.

Protein SUMOylation is thought to regulate cell migration and thus, we decided to assess whether blocking the SUMO pathway might also affect the invasive capacity of cancer cells. In accordance with this hypothesis, exposure to GA under conditions that did not affect cell viability (10 µM GA for 24 h) did inhibit breast cancer cell invasion and migration by more than 50% (Fig. 5A), as did genetic depletion of SUMO1 and UBC9 (Fig. 5B–E; Fig. S4B).

The GTPase RAC1 plays a crucial role in regulating cell migration, and it has been implicated in cancer cell progression. Likewise, increased expression of this GTPase is associated with a poor prognosis in different cancer types (Mack et al., 2011). Moreover, we previously showed that RAC1 is SUMOylated in response to stimuli that trigger cell migration (Castillo-Lluva et al., 2010). When we assessed whether the inhibition of cell invasion triggered by blocking the SUMO pathway involves a dampening of RAC1 activation, we found that GA did indeed inhibit RAC1 activity (RAC1-GTP) (Fig. 5F,G). Furthermore, breast cancer cells treated with GA acquired a non-migratory phenotype and the changes in the cytoskeleton of these cells resembled those observed in cells when RAC1 signalling was abrogated. Significantly, this effect was even more pronounced after 48 h in the presence of GA (Fig. 5H).

Consequently, we analysed whether the reduction in RAC1 activity and in the invasive capacity of breast cancer cells induced by GA was dependent on the reduction in RAC1 SUMOylation. To test this hypothesis, we first confirmed that GA reduced RAC1 SUMOylation (Fig. 5I). Subsequently, to validate the functional consequences of this effect, we transfected MDA-MB-231 breast cancer cells with plasmids expressing GFP, GFP-tagged wild-type RAC1 (GFP–RAC1) or a fusion protein containing GFP and a RAC1 chimaera that mimics RAC1 SUMOylation (GFP–SUMO1–RAC1) and, therefore, cannot be de-SUMOylated or be affected by GA. We then examined the effect of exposing these cells to GA. The expression of GFP–SUMO1–RAC1 but not that of wild-type RAC1 re-established the invasive capacity of these breast cancer cells upon exposure to GA (Fig. 5J–L), as well as their migratory phenotype as assessed by the capacity to form lamellipodia and ruffled structures in the proximity of the plasma membrane (Fig. 5M,N). By contrast, expression of GFP–SUMO1–RAC1 did not affect the decrease in cell viability evoked by GA (Fig. 5O). Together, these observations support the notion that inhibiting protein SUMOylation abrogates the migratory and invasive capacity of breast cancer cell lines by inhibiting RAC1 SUMOylation.

To investigate the relevance of our findings in vivo, we examined the effect of administering GA to athymic mice in which a tumour xenograft had been generated by subcutaneous injection of MDA-MB-231 cells into their right flank. Treatment with GA led to a dramatic reduction in the growth of these tumours (Fig. 6A; Fig. S4C). When the histology of these tumours was evaluated following haematoxylin-eosin (HE) staining, the tumours treated with GA contained areas rich in necrotic and apoptotic tissue, with a lower mitotic index than the control xenografts (Fig. 6B). Interestingly, PCNA staining revealed that there was a significant decrease in proliferation in the tumours exposed to GA relative to the controls (Fig. 6C,D), consistent with the lower mitotic index in the HE stained tissue. Moreover, immunofluorescence and immunoblotting on the final day of treatment showed that the levels of TRIB3 (Fig. 6E,H,I), LC3II (Fig. 6F,J,K) and cleaved C3 (Fig. 6G,L,M) were enhanced in these tumours when they had been treated with GA, supporting the idea that this compound also induces autophagy-mediated cell death in vivo. Finally, we investigated whether blockade of the SUMO pathway with GA affected the activation of RAC1 in these MDA-MB-231 cell-derived tumours. When we performed a pulldown assay of RAC1 in extracts from vehicle and GA-treated tumours, there was less active RAC1 in the tumours treated with GA (Fig. 6N,O).

Taken together, our findings offer strong support that the mechanism by which GA inhibits the SUMO pathway, and negatively regulates tumour survival and invasion in vitro, also operates in vivo.

Normal tissues control the production and release of signalling molecules in order to maintain their homeostasis and architecture. The dysregulation of such signals favours tumour progression, and the proliferation, survival and dissemination of cancer cells (Hanahan and Weinberg, 2011). SUMOylation is a PTM that alters the activity of proteins involved in different cellular processes and, thus, SUMOylation can regulate various biological processes (Johnson, 2004). Indeed, alterations in the SUMOylation–deSUMOylation equilibrium have been associated with various diseases, including cardiac pathologies, neurodegeneration and cancer (Yang et al., 2017). Here, we explored the importance of the SUMO1 pathway in controlling tumorigenesis by inhibiting SUMO1 expression genetically and pharmacologically. Our findings demonstrate that SUMO1 inhibition negatively regulates the tumorigenic properties of cancer cells via two different mechanisms: (1) stimulation of autophagy-mediated cancer cell death, and (2) inhibition of cancer cell invasiveness via regulation of RAC1 SUMOylation.

In our work, we used two pharmacological inhibitors of protein SUMOylation acting through different mechanisms: C15:1 GA and 2D-08. GA is a phenolic acid obtained from leaves, nuts and the external seed coat of G. biloba, which exert several pharmacological effects including inhibition of tumour growth in animal models of cancer (Baek et al., 2017; Li et al., 2017; Liu et al., 2018b). Specifically, we used the C15:1 monomer that inhibits protein SUMOylation by blocking the E1–SUMO1 interaction (Fukuda et al., 2009). We show that, as with other molecular species (Li et al., 2017; Liu et al., 2018a,b), the viability of breast and prostate cancer cells was compromised by GA in a dose-dependent manner. However, exposure to GA (10 µM) for 24 h impaired the formation of the SUMO1 complexes without affecting cell viability. Thus, in these conditions, we could investigate the implication of SUMO inhibition in tumorigenesis.

In our study, we also used the SUMO E2 inhibitor 2-D08, which has been shown to block SUMOylation (Choi et al., 2018; Kim et al., 2014), although GA proved to be a more potent SUMO inhibitor than 2-D08, producing a 40% reduction in SUMO1-complexes after 24 h while 2-D08 failed to reduce SUMOylation in that time. However, after a 48 h exposure to 2-D08, similar inhibition of SUMOylation was evident as that produced by GA and a similar degree of autophagy-induced cell death was produced by both agents.

Findings presented in this work demonstrate by using different experimental approaches that SUMO1 inhibition induces autophagy-mediated cancer cell death. Thus, inhibition of SUMO1 induces the upregulation of several autophagy related genes such as ATG1, ATG5, ATG7, BECN1 or AMBRA1 at early time points. Moreover, the genetic depletion of some of this key autophagy regulatory genes (or the pharmacological inhibition of autophagy, E64d and pepstatin-A) prevents SUMO1 blockade-induced apoptosis and cell death. It is worth noting that, unlike genetic inhibition of SUMO1, the depletion of SUMO2/3 does not induce autophagy-mediated cell death, suggesting that the effects observed in this work take place specifically through the blockade of SUMO1 conjugation.

GA and 2D-08 has been shown to inhibit the SUMO pathways acting on the E1–SUMO and E2–SUMO enzymes respectively. However, whether these compounds, and specifically GA, may act on other targets apart from these SUMO-regulatory enzymes is unknown. For example, different studies have shown that treatment with GA inhibits lipogenesis (Ma et al., 2015), and the EGF (Li et al., 2017) and TGF-β (Baek et al., 2017) signalling pathways, among others. However, none of these studies investigated whether the changes observed on those signalling pathways relied on the modulation of protein SUMOylation. Our results clearly demonstrate that treatment with GA and 2-D08 inhibits protein SUMOylation in the cellular models employed in this study; however, it cannot be ruled out that GA (or 2-D08) may produce their effects at least in part acting independently of the regulation of protein SUMOylation.

TRIB3 is a member of the Tribbles family of pseudokinases that has been implicated in autophagy-mediated cancer cell death after administration of different anti-cancer agents (Salazar et al., 2009a,b). In our work, we demonstrate that pharmacological or genetic inhibition of the SUMO pathway induces autophagy, apoptosis and cell death in a TRIB3-dependent manner. Moreover, we found that TRIB3, as well as other markers of the autophagy-mediated cell death pathway, are upregulated in vivo upon GA administration to mice bearing breast cancer xenografts. However, TRIB3 depletion partially rescued the decrease in MDA-MB-231 viability evoked by GA treatment, indicating that there could be additional (TRIB3-independent) mechanisms by which GA regulates cancer cell death.

RAC1 plays an important role in controlling the cytoskeletal reorganization associated with cell motility. In a cancer context, the enhanced activation of this protein has been implicated in metastasis and invasiveness. Furthermore, RAC1 is SUMOylated by SUMO1 and this modification is required to maintain the activity of the protein during cell migration (Castillo-Lluva et al., 2010). Here, we demonstrate that RAC1 SUMOylation is important for optimal breast tumour cell migration and invasion but not for other RAC1 functions, such as proliferation and survival. GA inhibits the migration of various tumour cell types (Baek et al., 2017; Hamdoun and Efferth, 2017; Yang et al., 2014), and we show that this effect on cell invasion is related to the inhibition of SUMO conjugation.

Overall, the data presented here enhance our understanding of the mechanisms by which protein SUMOylation regulates cancer cell biology. Specifically, we found that inhibiting protein SUMOylation induces autophagic cell death through the upregulation of the pseudokinase TRIB3, and it impairs cancer cell invasiveness by inhibiting activation of the small GTPase RAC1. Blocking the SUMO pathway would reduce RAC1 SUMOylation, thereby diminishing the amount of active RAC1 available to drive the cell migration and invasion programme. Nevertheless, further studies will be required to fully understand the mechanisms involved in the induction of autophagy following the inhibition of SUMOylation. In any case, our findings do support the notion that blocking protein SUMOylation can be explored as a potential therapeutic strategy to fight cancer.

Ginkgolic acid C15:1 (#02580585), 2′,3′,4′-trihydroxy-flavone, 2-(2,3,4-trihydroxyphenyl)-4H-1-Benzopyran-4-one (2-D08; #SML1052), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; #M5655), Crystal Violet (#C3886), and the E64d (10 µmol/l; #E8640) and Pepstatin-A (10 µg/ml; #P5318) inhibitors were all purchased from Sigma-Aldrich (St Louis, MO). HeLa cells were transfected using LT1 (Mirus Bio, Madison WI; #MIR 2304) and MDA-MB-231 cells with Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA; #11668027). The SUMO1 siRNA (#sc-29498), UBC9 siRNA (#sc-36773), SUMO2/3 siRNA (#sc-37167), BECLIN1 siRNA (sc-29797), AMBRA1 siRNA (sc-96257) and TRIB3 shRNA lentiviral particles (#sc-44426-V) were obtained from Santa Cruz Biotechnology (Heidelberg, Germany). SUMO1 siRNA #2 was from Sigma-Aldrich (#EHU106621). For siRNA transfection, DharmaFECT 2001 was used (Dharmacon, Lafayette, CO; #T2001).

Proteins were analysed in western blots as described previously (Castillo-Lluva et al., 2015). In brief, proteins were extracted from tumours frozen in RIPA buffer [150 mM NaCl, 1% (v/v) NP40, 50 mM Tris-HCl pH 8.0, 0.1% (v/v) SDS, 1 mM EDTA, 0.5% (w/v) deoxycholate], while proteins from cell lines were extracted in TNES buffer [100 mM NaCl, 1% (v/v) NP40, 50 mM Tris-HCl pH 7.6, 20 mM EDTA] containing protease and phosphatase inhibitor cocktails (Sigma-Aldrich; #P8340). The proteins recovered were quantified using the BCA Protein Assay Kit (Thermo Fisher Scientific; #23228), resolved by SDS-PAGE on 4–15% gradient gels (purchased from Bio-Rad, Berkeley, CA; #456-8085) and transferred to polyvinylidene difluoride membranes (Immobilon-P, Millipore, Burlington, MA). The membranes were then probed with the following primary antibodies: anti-cJUN (1:500; H72, Santa Cruz Biotechnology; #sc-1694), anti-Bax (1:1000; Cell Signaling Technology; #2772S), anti-TRIB3 (1:1000; Abcam, Cambridge, UK; #ab75846), anti-tubulin (1:1000; DM1A, Sigma-Aldrich; #T6199), anti-RAC (1:1000; clone 102; BD Biosciences, Franklin Lakes, NJ), anti-LC3 (1:1000, Sigma-Aldrich; #L7543), anti-cleaved caspase-3 (1:1000, Cell Signaling Technology; #9662), anti-GAPDH (1:1000, Sigma-Aldrich; #G8795), anti-cleaved PARP (1:1000, Cell Signaling Technology; #D214), anti-UBC9 (1:1000, Abcam; #ab75854), anti-SUMO1 and anti-SUMO2/3 (provided by Ron Hay's laboratory, School of Life Sciences, University of Dundee, Dundee, UK), anti-GFP (Sigma-Aldrich; #SAB1305545), and anti-ATG5 (Cell Signaling Technology; #12994S). Antibody binding was detected with horseradish peroxidase (HRP)-conjugated anti-mouse, anti-rabbit or anti-sheep secondary antibodies (1:10,000 dilution, Bio-Rad), and visualized by enhanced chemiluminescence (Bio-Rad). The images were obtained with the ImageQuant LAS 500 chemiluminescence CCD camera (GE Healthcare Life Sciences, Chicago, Illinois, USA).

MDA-MB-231 and MCF7 breast cancer cells (ATCC) were grown in complete DMEM, containing 4.5 g/l glucose and L-glutamine (Sigma-Aldrich). LnCaP and 22Rv1 prostate cancer cells, and BT474 breast cancer cells (ATCC), were grown in RPMI medium (Sigma-Aldrich). RasV¹²/T-large antigen-transformed ATG5^+/+ and ATG5^−/− cells (Salazar et al., 2009b) were grown in in DMEM. In all cases, the medium was supplemented with 56 IU/ml penicillin, 56 mg/l streptomycin (Invitrogen, Carlsbad, CA) and 10% fetal bovine serum (FBS, LINUS #16sV30180.03), and the cells were maintained at 37°C in a humid atmosphere containing 5% CO₂. All cells were routinely tested for mycoplasma contamination. Cell lines were authenticated at the Genomics Core Facility (Instituto de Investigaciones Biomédicas ‘Alberto Sols’ CSIC-UAM, Madrid, Spain) using the STR PROFILE DATA, STR amplification kit, GenePrintR 10 System (Promega), STR profile analysis software GeneMapper^® v3.7 (Life Technologies) and the Genomic Analyzer System ABI 3130 XL (Applied Biosystems). Cell viability was determined in MTT assays following the manufacturer's instructions (Sigma-Aldrich; #M5655). For transfection experiments with SUMO1 and control siRNA, 10⁶ cells were transfected using the DharmaFECT 1 Transfection Reagent (Dharmacon; #T-2001) and incubated for up to 72 h before setting up the experiment. Infection with TRIB3 shRNA human lentiviral particles was performed by using a pool of concentrated transduction-ready viral particles containing three target specific shRNAs (or three non-targeted control shRNA) constructs (Santa Cruz Biotechnology). After the infection of human cancer cell lines, stably silenced (or control shRNA transduced) cells were selected.

RAC1 activation was determined using a RAC1 activation pulldown assay (Thermo Fisher Scientific; #16118) following the manufacturer's instructions. Cells were treated with 10 µM of GA C15:1 for 24 h before RAC1 activity was determined.

Cell migration was assayed in Boyden chambers (8.0 μm pore-size polyethylene terephthalate membrane with a cell-culture insert: VWR, Radnor, PA; #VWRI734-2744). The cells were trypsinized and counted, and cell suspensions containing 5×10⁴–10×10⁴ cells in 300 μl of serum-free medium were added to the upper chamber, with 500 μl of the appropriate medium added to the lower chamber. The transwell inserts were incubated for 24 h at 37°C, and the cells on the inside of the transwell inserts were removed with a cotton swab, whereas those on the underside of the insert were fixed and stained. Photographs were taken of five random fields and the cells were counted to calculate the proportion of transmigrated cells. Boyden chambers with Matrigel (Cell Biolabs, San Diego, CA; #CBA-110) were used to assay cell invasion.

MDA-MB-231 cells were plated and grown on glass coverslips for 24 h. The cells were fixed for 10 min in PBS with 4% paraformaldehyde at room temperature and the coverslips were then stored in PBS with 0.05% azide at 4°C until the cells were immunostained. Briefly, the cells were permeabilized by incubating for 3 min at 4°C in PBS containing 0.5% Triton X-100 and after that, the coverslips were incubated overnight at 4°C with a 1:500 dilution of the primary antibodies against Ki67 (Thermo Fisher Scientific, #PA1-21520), LC3 (Sigma-Aldrich; #L7543) or 1:50 of cleaved caspase-3 (Asp175; Cell Signaling; #9661). After rinsing thoroughly, the coverslips were incubated with Alexa Fluor 488 or 594 goat anti-mouse-IgG or anti-rabbit-IgG secondary antibodies (Invitrogen), counterstained with DAPI for 30 s, and then mounted on glass slides with DAPI using the Gold antifade reagent (Invitrogen; #P36935). Immunofluorescence was analysed under a Zeiss microscope (Zeiss Axioplan 2).

MDA-MB-231 apoptotic cells were quantified using an Annexin V-fluorescein isothiocyanate (FITC)/propidium iodide (PI) apoptosis detection kit (Sigma-Aldrich; #APOAF). Cells were seeded at a density of 10⁶ cells in 10 cm dishes. After treatment with 10 µM of GA for 48 h in serum-free medium, cells were harvested and Annexin-V-FITC/PI labelling was performed according to the manufacturer's instructions. The stained cells were analysed with a flow cytometer (FACSCalibur, BD Biosciences).

His6–SUMO–1-binding proteins were purified from cell lysates using nickel-nitrilotriacetic acid agarose (QIAGEN, Hilden, Germany; #30230), as described previously (Castillo-Lluva et al., 2010).

Tumours were generated in nude mice (Athymic nude-Foxn1 6 weeks old, ENVIGO RMS, Barcelona, Spain) by subcutaneous injection of 10×10⁶ MDA-MB-231 cells in PBS supplemented with 0.1% glucose. When the tumours had reached an average volume of 200 mm³, the animals were randomly assigned to different treatment groups and injected peritumourally for 18 days with GA (10 mg/kg body weight per day) or the vehicle alone (ethanol) in 100 µl of PBS, supplemented with 5 mg/ml defatted and dialyzed BSA. Tumour growth was determined with digital callipers and the tumour volume was estimated on each day using the formula: tumour volume=length×width²×0.5. We obtained the tumour growth rate after transforming the data logarithmically, and we estimated a linear regression curve for each tumour, evaluating the means of the slopes of these lines for each group.

This research was carried out in accordance with the regulatory ethical standards, and in compliance with national and international guidelines, and it was approved by the authors' institutional review board.

Unless otherwise indicated, the data are expressed as the mean±standard error of the mean (s.e.m.), and evaluated with the Mann–Whitney U-test or t-test for two groups, with the Kruskal–Wallis test (followed by Dunn's multiple comparison post-test) or by ANOVA (followed by Turkey's multiple comparisons test) for more than two groups. The percentage of positive cells was evaluated using the χ-squared test and the tumour volumes in the mice were compared using a multiple t-test. For all the analyses, P-values ≤0.05 were considered statistically significant.

We thank Prof. Angeliki Malliri for providing reagents for the SUMOylation experiments and Ron Hay for the gift of the anti-SUMO1 antibody. We would also like to thank Dr Jesús Pérez-Losada for scientific support and all the Biochemistry and Molecular Biology department for personal and technical support, especially Manuel Guzman.