Sorafenib induces macropinocytosis in human HCC tissues and HCC cells
To determine whether sorafenib induces macropinocytosis in HCC, we assessed alterations in macropinocytic uptake in tumor tissues obtained from treatment-naïve HCC patients. Clinical diagnostics were performed, including immunofluorescence staining for the hepatocellular marker HepPar-1 (Table S1 and Fig. S1A). Macropinosomes were detected using tetramethylrhodamine-labeled high-molecular-mass dextran (TMR-dextran), an established marker of macropinocytosis [19]. Notably, TMR-dextran uptake by human HCC tissues treated with sorafenib was significantly higher than by tissues treated with vehicle (Fig. 1A, B and Fig. S1B). Consistent with a role for PAK1 in regulating inducible macropinocytosis [20], we observed a marked increase in phosphorylated PAK1 in sorafenib-treated human HCC tissues (Fig. 1A, B and Fig. S1B). Sorafenib-induced robust macropinocytic TMR-dextran uptake was further confirmed in multiple HCC cell lines including SK-Hep1, Huh7, PLC/PRF/5, and Hep3B cells; this was attenuated significantly by treatment with a pharmacological blocker of macropinocytosis, 5-(N-ethyl-N-isopropyl)-amiloride (EIPA), which inhibits plasma membrane Na+/H+ exchangers (NHEs) (Fig. 1C, D and S1C). Increased uptake of macropinocytic TMR-dextran induced by sorafenib was also observed in renal cancer cells Caki-1 and ACHN, suggesting that sorafenib-induced macropinocytosis occurs in other cancer cells (Fig. S1C). To determine whether internalized albumin is degraded intracellularly, we used boron-dipyrromethene (BODIPY)-conjugated BSA (DQ-BSA), which is taken up by macropinocytosis and fluoresces after lysosomal degradation [21]. Indeed, sorafenib-treated HCC and renal cancer cells showed increased DQ-BSA fluorescence, which was attenuated by EIPA (Fig. 1E, F and Fig. S1D). Furthermore, sorafenib-induced dextran or DQ-BSA uptake was attenuated at 4 °C, or by treatment with NH4Cl (an uncoupler of lysosomal acidification), respectively, which also supports the notion that sorafenib-induced uptake and degradation of extracellular proteins is mediated by macropinocytosis (Fig. S1E and F).
Consistent with the role of RAC1, PAK1, and CDC42 in regulating inducible macropinocytosis [22], analysis of the TCGA dataset revealed higher expression of RAC1, PAK1, and CDC42 in HCC tissues than in non-tumor lesions (Fig. 1G). High levels of RAC1, PAK1, and CDC42 expression correlated with shorter overall survival (Fig. 1H). When these clinical results are considered alongside induction of macropinocytosis by sorafenib, it appears that levels of macropinocytosis in HCC predict prognosis and, furthermore, that induction of macropinocytosis may be associated with the outcome of sorafenib treatment.
Macropinocytosis-mediated acquisition of cysteine enables HCC cells to acquire resistance to sorafenib-induced ferroptosis
The findings described above prompted further investigations to determine the effect of macropinocytosis induction on sorafenib-induced ferroptosis. Consistent with previous results [10, 11], we found that genes up-regulated in response to sorafenib were enriched in a ferroptosis-related gene signature from human fibrosarcoma cells treated with the selective xCT inhibitor erastin (Fig. 2A) [23]. Analysis of tumor tissues from HCC patients revealed increased staining of 4-HNE, a byproduct of lipid peroxidation, after sorafenib treatment (Fig. 2B and Fig. S2A). Given that sorafenib inhibits SLC7A11 directly and consequently abolishes glutathione (GSH) biosynthesis and induces ferroptosis [24], we monitored lipid ROS accumulation, which is a hallmark of ferroptosis, using C11-BODIPY. As expected, sorafenib increased lipid ROS accumulation in HCC cells, whereas supplementation with GSH, or β-mercaptoethanol (β-ME) which promotes cystine uptake in an xCT-independent manner [25], prevented sorafenib-induced lipid ROS accumulation (Fig. 2C and Fig.S2B). Because lysosomal catabolism of extracellular albumin taken up by macropinocytosis plays a role in maintaining intracellular amino acid availability, we reasoned that uptake of extracellular albumin might prevent ferroptosis in the presence of sorafenib. To test this idea, we assessed lipid ROS accumulation in sorafenib-treated HCC cells with and without albumin supplementation. We found that addition of BSA reduced sorafenib-induced ferroptosis, as evidenced by attenuated lipid oxidation and 4-HNE staining intensity, whereas the inhibitory effect of BSA on sorafenib-induced ferroptosis was suppressed by EIPA (Fig. 2D and Fig. S2C). EIPA alone did not increase accumulation of lipid ROS, suggesting that EIPA increased ferroptosis by inhibiting sorafenib-induced macropinocytosis (Fig. S2D and E). In line with this finding, mRNA sequencing data revealed no upregulation of ferroptosis-related genes in sorafenib-treated HCC cells in the presence of BSA; BSA had no effect when macropinocytosis was inhibited by EIPA (Fig. 2E). Consistent with previous results showing that sorafenib induced ferroptosis by depletion of cellular cysteine, treatment with sorafenib decreased intracellular cysteine levels in HCC cells (Fig. 2F). However, in the presence of BSA, HCC cells recovered from sorafenib-induced cysteine depletion significantly, which was reversed by treatment with EIPA or bafilomycin (Fig. 2F). We also observed that intracellular GSH levels correlated with cysteine levels in the presence or absence of BSA and a macropinocytosis inhibitor (Fig. 2G). Given that cysteine is used to synthesize GSH required to mitigate ferroptosis, we examined whether inhibiting GSH synthesis or lysosomal degradation of BSA abolishes the protective effect of macropinocytosis on sorafenib-induced ferroptosis. Indeed, sorafenib-induced ferroptosis occurred upon treatment with the GSH synthesis inhibitors buthionine sulfoximine (BSO) or bafilomycin in the presence of BSA (Fig. S2F), indicating that macropinocytosis plays a role in preserving intracellular cysteine levels and preventing ferroptosis in sorafenib-treated HCC cells. Given the contribution of the transsulfuration pathway to intracellular cysteine levels, we further analyzed expression of cystathionine-β-synthase (CBS) and cystathionine γ-lyases (CTH), two enzymes involved in the transsulfuration pathway. We also examined expression of genes involved in the methionine cycle in sorafenib-treated HCC cells. Expression of CBS and CTH increased in response to sorafenib, whereas that of genes involved in methionine cycle were regulated inconsistently (Fig. S2G), which is discrepant with our finding that sorafenib depletes intracellular cysteine. Furthermore, in contrast to depletion of intracellular cysteine by EIPA in the presence of BSA, RNA-seq analyses revealed that expression of CBS and CTH in response to EIPA tended to remain high (Fig. S2H). Notably, although CBS was upregulated by sorafenib, it did not prevent lipid ROS accumulation, nor did silencing of CBS amplify sorafenib-induced lipid ROS accumulation significantly (Fig. 2H, I and Fig. S2I). Taken together, these data reveal that the transsulfuration pathway cannot compensate for the sorafenib-induced deficit of intracellular cysteine, and that macropinocytosis plays a pivotal role in preserving intracellular cysteine and GSH levels, which support the fitness of HCC cells and allows them to escape the deleterious effects of ferroptosis.
Sorafenib-induced mitochondrial dysfunction drives macropinocytosis by activating PI3K-RAC1-PAK1 signaling
To investigate the mechanisms by which sorafenib drives macropinocytosis in HCC cells, we investigated whether key regulators of macropinocytosis (such as PI3K, RAS, and WNT) increased in HCC cells exposed to sorafenib [3]. Notably, RNA-seq profiling of sorafenib-treated HCC cells revealed that genes activated by PI3K were highly enriched (Fig. 3A). By contrast, RAS and WNT signaling was not altered significantly upon sorafenib treatment (Fig. S3A). PI3K affects RAC1 activity by regulating the activity of PIP3-dependent RAC1 guanine nucleotide exchange factors, and activated RAC1 binds to and stimulates the kinase activity of PAK1 via intramolecular inhibition and consequent autophosphorylation, which drives macropinocytosis through instigation of macropinocytic induction mechanisms [26,27,28]. Indeed, we found that sorafenib-treated cells increased PI3K activity by increasing expression of the phosphorylated PIP3-dependent Ser/Thr kinase AKT and found it increased levels of RAC1-GTP and phosphorylation of PAK1 (Fig. 3B).
To further elucidate the mechanism responsible for sorafenib-induced macropinocytosis, we firstly examined whether xCT inhibition is implicated in macropinocytosis of HCC cells. As shown in Fig. S3B, erastin did not increase dextran uptake in HCC cells, suggesting xCT-independent macropinocytosis by sorafenib. Considering that sorafenib targets mitochondrial biogenesis, we next examined the impact of sorafenib-induced mitochondrial dysfunction on macropinocytosis in HCC cells. As expected, sorafenib decreased mitochondrial oxygen consumption by HCC cells (Fig. 3C and Fig. S3C). Since sorafenib induces mitochondrial dysfunction by inhibiting complex II/III in the electron transporter chain, as well as ATP synthase [29], we investigated whether co-treatment with oligomycin and antimycin increases expression of major kinases required for macropinocytosis induction in HCC cells. In parallel with previous findings, we observed that HCC cells treated with oligomycin and antimycin exhibited increased phosphorylation of AKT and PAK1, accompanied by increased TMR-dextran uptake and DQ-BSA fluorescent puncta (Fig. 3D and Fig. S3D, E). After demonstrating that mitochondrial dysfunction is implicated in macropinocytosis induction in sorafenib-treated HCC cells, we investigated the effects of PI3K inhibition on sorafenib-induced macropinocytosis. The results showed that the sorafenib-induced increase in RAC1-GTP levels was strongly suppressed by treatment with a PI3K inhibitor, suggesting PI3K-dependent RAC1 regulation in HCC cells (Fig. 3E). Furthermore, sorafenib increased PAK1 phosphorylation and TMR-dextran or DQ-BSA uptake, which were strongly suppressed by treatment with PI3K or RAC1 inhibitors (Fig. 3F, G and H). Finally, sorafenib-induced macropinocytosis in HCC cells was abrogated by a PAK1 inhibitor (Fig. S3F). Given that sorafenib activates AMPK indirectly by inhibiting mitochondrial function [30], and the requirement for AMPK to support RAC1 activation and macropinosome formation [15, 17], we also investigated the roles of AMPK in sorafenib-induced macropinocytosis. Treatment with sorafenib, or co-treatment with oligomycin and antimycin, increased phosphorylation of AMPK, while an AMPK inhibitor abrogated both sorafenib-induced RAC1 activity (as measured by inhibition of PAK1 phosphorylation) and induction of macropinocytosis (Fig. 3B, D, F, G and H).
Inhibiting macropinocytosis with amiloride recovers sorafenib-induced ferroptosis in HCC
After demonstrating that macropinocytosis protects sorafenib-treated HCC cells from ferroptosis, we investigated whether blocking macropinocytosis with amiloride, a clinically feasible NHE inhibitor, potentiates the antitumor effect of sorafenib [31]. Indeed, sorafenib-induced macropinocytosis was attenuated by treatment with amiloride (Fig. 4A and B). Furthermore, blocking macropinocytosis with amiloride potentiated ferroptosis, as evidenced by increased lipid oxidation and 4-HNE staining intensity; these were prevented by cotreatment with ferroptosis inhibitors ferrostatin-1 (a lipophilic antioxidant) or deferoxamine (an iron chelator) (Fig. 4C, D and Fig. S4A, B). Previous results show compensatory transcriptional upregulation of SLC7A11 upon inhibition of system xc − [7, 32]; thus, we next examined alteration of SLC7A11 expression in response to amiloride in the presence of sorafenib. Consistent with previous studies, we found substantial upregulation of SLC7A11 in sorafenib-treated HCC cells (Fig. 4E). Supplementation with BSA reduced sorafenib-induced upregulation of SLC7A11, which was reversed when sorafenib was combined with amiloride (Fig. 4E). These findings, combined with results shown in Fig. 4C and D, suggested that transcriptional upregulation of SLC7A11 is not sufficient to prevent ferroptosis in sorafenib-treated HCC cells. Next, we investigated the contribution of ferroptosis in sorafenib-induced downregulation of HCC cell numbers. As shown in Fig. S4C, neither apoptosis nor necroptosis signaling was activated in sorafenib-treated HCC cells, consistent with previous findings [24]. Rather, sorafenib-induced decreases in HCC cell numbers were reversed significantly by co-treatment with ferroptosis inhibitors ferrostatin-1, trolox, and GSH, but not by treatment with the apoptosis inhibitor Z-VAD-FMK or the necroptosis inhibitor necrostatin-1 (Fig. S4D and E). In accordance with these results, recovery of HCC cell numbers in the presence of BSA was abolished when macropinocytosis was inhibited by amiloride or EIPA, which do not affect apoptosis or necroptosis signaling significantly (Fig. 4F and Fig. S4F, G). A clonogenic assay also confirmed that macropinocytosis renders HCC cells resistant to sorafenib treatment (Fig. 4G).
Inhibiting macropinocytosis with amiloride increases the antitumor efficacy of sorafenib
The findings described above suggest that macropinocytosis makes HCC cells resistant to sorafenib-induced ferroptosis; therefore, we investigated whether inhibiting macropinocytosis increases the antitumor activity of sorafenib in vivo. Although treatment of SK-Hep1- or Huh7-bearing mice with sorafenib led to a marked reduction in tumor mass, it did not completely inhibit tumor growth (Fig. 5A, B and Fig. S5A–D). However, simultaneous treatment with sorafenib and amiloride suppressed tumor growth almost completely, without any differences in body weight between the groups (Fig. 5A, B and Fig. S5A–D). In parallel with tumor growth, morphological analysis revealed that combined treatment with sorafenib and amiloride markedly increased the number of necrotic lesions with an unusual cellular phenotype, which was characterized by ballooned cells showing cytoplasmic vacuolization, compared with sorafenib treatment alone (Fig. 5C, D). Further analyses of xenograft tumor tissues by immunofluorescence revealed that sorafenib alone increased 4-HNE staining, but not cleaved caspase-3 staining (Fig. 5C, D and Fig. S5E, F). Importantly, combined treatment with sorafenib and amiloride increased 4-HNE staining significantly without altering cleaved caspase-3 staining in tumor tissues from SK-Hep1- and Huh7-bearing mice (Fig. 5C and D and Fig. S5E, F), indicating that ferroptosis is responsible for the increased therapeutic effect of sorafenib. We further investigated the relevance of macropinocytosis to reduced tumor growth induced by co-treatment with sorafenib and amiloride. Consistent with the in vitro results, we found that combined treatment with amiloride attenuated sorafenib-induced dextran uptake and staining of phosphorylated AKT and PAK1 in HCC tissues from SK-Hep1 or Huh7 xenograft tumors (Fig. 5E and Fig. S5G), which highlights the important role of inhibiting macropinocytosis in increasing the susceptibility of HCC to sorafenib. The combined effects of sorafenib and amiloride on tumor growth, the area of necrosis, and the levels of phosphorylated AKT, PAK1, and 4-HNE after blocking macropinocytosis were further verified in orthotopic RIL-175 tumors transplanted into C57BL/6 mice (Fig. 5F, G and Fig. S5H, I and J).
Sorafenib-resistant HCC cells exhibit high levels of macropinocytosis, and inhibiting macropinocytosis sensitizes resistant tumors to sorafenib
The role of macropinocytosis in triggering resistance to sorafenib-induced ferroptosis provided us with a rationale to use macropinocytosis inhibition as a method of resensitizing sorafenib-resistant tumors. To this end, sorafenib-resistant SK-Hep1 and Huh7 cells were established via exposure to gradually increasing concentrations of sorafenib (up to 10 μM) for ~ 5 months (Fig. 6A). Interestingly, levels of phosphorylated PAK1 were higher in sorafenib-resistant SK-Hep1 and Huh7 cells than in parental cells (Fig. 6B). Moreover, sorafenib-resistant HCC cells exhibited higher dextran uptake and more DQ-BSA fluorescent puncta than parental cells in the presence or absence of sorafenib; these phenomena were abolished by co-treatment with EIPA or amiloride (Fig. 6C and D), suggesting the potential role of macropinocytosis in acquired sorafenib resistance. Finally, we investigated in vivo co-operativity between amiloride and sorafenib in mice bearing sorafenib-resistant HCC cells. Sorafenib or amiloride alone did not reduce tumor growth, whereas combined treatment with sorafenib and amiloride led to a marked reduction in growth of sorafenib-resistant tumors, with no differences in body weight between groups (Fig. 6E and Fig. S6A and B). Consistent with this, sorafenib alone did not induce necrotic lesions in sorafenib-resistant tumors, whereas it markedly increased the number of necrotic lesions when combined with amiloride (Fig. 6F and Fig. S6C). Increased 4-HNE staining further confirmed that amiloride enhanced the therapeutic effects of sorafenib against HCC by increasing ferroptotic cell death (Fig. 6F and Fig. S6D). Collectively, these findings illustrate the potential therapeutic benefits of combined treatment of HCC with macropinocytosis inhibitors and sorafenib.