Palmitate-induced insulin resistance causes actin filament stiffness and GLUT4 mis-sorting without altered Akt signalling Free

Skeletal muscle insulin resistance occurs as a consequence of nutrient excess and is a major contributor to type 2 diabetes (Schenk et al., 2008). Skeletal muscle takes up over 70% dietary glucose from the circulation (Thiebaud et al., 1982) and is hence a key determinant of glycemia. Glucose enters skeletal muscle via the GLUT4 (also known as SLC2A4) transporter, which undergoes redistribution from intracellular storage endomembranes to the muscle cell surface in response to insulin (Jaldin-Fincati et al., 2017). GLUT4 translocation is an integrated process initiated by insulin receptor-mediated phosphorylation and consequent phosphorylation of insulin receptor substrate-1 (IRS-1) and activation of class I phosphatidylinositol-3-kinase (PI3K). Downstream signalling bifurcates towards activation of Akt through to phosphorylation of Akt substrate of 160 kDa (AS160; also known as TBC1D4), which activates Rab GTPases (such as Rab8A, Rab10 and Rab13), in parallel with mechanical/structural events governed by the small GTPase Rac1, resulting in cortical actin cytoskeleton remodelling. Together, these two axes converge to rapidly deliver GLUT4 from intracellular storage vesicles to the plasma membrane (PM) where it facilitates glucose uptake (Klip et al., 2014).

Insulin resistance is almost invariably associated with reduced glucose uptake into muscle (Shulman, 2000) and, in instances where it has been measured, the insulin-dependent gain in surface GLUT4 is diminished (Tremblay et al., 2001). This failure of GLUT4 translocation during insulin resistance has been largely attributed to a partial deficit in insulin receptor signalling to Akt, but data from our laboratory and others indicate that even a very low level of Akt activation is sufficient to support GLUT4 translocation (Tan et al., 2012; Wang et al., 1999). Moreover, diminished insulin-stimulated GLUT4 translocation can proceed in conditions that preserve insulin-stimulated Akt phosphorylation (Kruszynska et al., 2002; Raun et al., 2018; Tan et al., 2015). Conversely, defects in GLUT4 translocation steps that are independent of Akt signalling, such as those involving Rac family small GTPase 1 (Rac1) and p21-activating kinase 1 (PAK1), also accompany insulin resistance, and can on their own affect insulin-stimulated glucose uptake (Merz et al., 2022; Sylow et al., 2013). It emerges that different alterations are observed in the various conditions causing or associated with insulin resistance, but only selective aspects of GLUT4 translocation have been examined in each case.

Palmitate (PA, C:16:0) is the most prevalent saturated fatty acid in the Western diet and numerous studies show that PA provokes skeletal muscle insulin resistance in skeletal muscle cell lines (Dimopoulos et al., 2006; Yang et al., 2013). Although it is documented that insulin-stimulated GLUT4 translocation fails in response to insulin after PA treatment, the specific signals or steps leading to GLUT4 translocation that are responsible for this defect have not been identified. In particular, whether saturated fats affect cellular parameters that render cells unresponsive to insulin, as opposed to affecting insulin signalling, requires thorough examination.

To this end, we have employed a reductive ‘disease-in-a-dish’ cell culture model of rat-derived skeletal muscle cells that endogenously possess all the machinery required for insulin-responsive GLUT4 translocation. Our findings reveal that PA confers diminish insulin-stimulated GLUT4 translocation despite preserved insulin signaling to Akt and AS160. Instead, PA interrupts the intracellular itinerary of GLUT4, causing mislocalization and expansion of the normally compact perinuclear storage compartment, impairing intracellular GLUT4 sorting to the insulin-responsive syntaxin-6 (STX6)-bearing subcompartment. Simultaneously, PA impairs Rac1-dependent actin remodelling and signaling to its downstream effectors [PAK1, PAK2, cofilin-1 and actin-related protein 2/3 complex subunit 2 (ArpC2)]. These alterations were brought about by independent upstream pathways: PA-elicited mislocalization of GLUT4, which involved palmitoylation [mitigated by 2-bromopalmitate (2-BP)], whereas defects in actin remodelling were caused by endoplasmic reticulum (ER) stress [and were mitigated by co-treatment with the ER stress inhibitor 4-phenylbutyric acid (4-PBA)]. However, 2-BP and 4-PBA alone or jointly were insufficient to restore GLUT4 translocation, indicating that additional or distinct upstream inputs contribute to dampening GLUT4 translocation. Nonetheless, the changes can be prevented by co-administration of the unsaturated fat palmitoleate (PO, C:16:1). On its own, PO did not cause insulin resistance, and effectively ameliorated the actions of PA on GLUT4 translocation, intracellular localization and actin remodelling.

Cultures of L6 skeletal muscle myoblasts stably expressing Myc-tagged GLUT4 (GLUT4myc) and the human insulin receptor (herein referred to as L6-GLUT4myc-hIR) were used for this study. GLUT4myc expression allows for detection of GLUT4 transporters inserted into the PM by antibody labelling of the exofacial Myc epitope through rapid and sensitive densitometric or fluorescence assays (Jaldin-Fincati et al., 2018; Wang et al., 1998), and insulin receptor overexpression confers sensitivity to insulin at physiological doses as low as 0.1 nM (Fig. S1A).

We first evaluated GLUT4 translocation after an 18-h treatment with increasing doses of BSA-conjugated PA (0–0.350 mM) to determine the lowest dose of PA required to cause insulin resistance of GLUT4 translocation. After PA treatment, myoblasts were serum-starved for 3 h and stimulated with 0.1 nM insulin for 15 min. Although insulin-stimulated GLUT4 translocation proceeded normally in controls treated with BSA, PA caused a dose-dependent decrease in the insulin-stimulated gain in surface GLUT4, as measured by densitometry (Fig. 1A) and immunofluorescence (Fig. 1B). At 0.350 mM PA, insulin failed to significantly stimulate GLUT4 translocation (relative to the unstimulated control) and did not affect cell viability (as measured by LDH release, Fig. S1B) or total GLUT4myc expression (Fig. S1C).

Insulin triggers a signalling cascade that is often typified by phosphorylation of the serine/threonine kinase Akt2 (hereafter Akt), and partially reduced Akt phosphorylation is observed in numerous insulin resistant states (reviewed in Abdul-Ghani and Defronzo, 2010). Although partially reduced Akt phosphorylation is often cited as a causal determinant of impaired GLUT4 translocation, 5–10% of maximal Akt phosphorylation is sufficient to sustain full insulin action (Wang et al., 1999). We therefore examined the status of insulin-stimulated Akt phosphorylation in the specific setting of PA-induced impairment of GLUT4 translocation. Under conditions that progressively reduced GLUT4 translocation (0–350 mM PA, 18 h), the robust insulin-stimulated phosphorylation of Akt at threonine 308 and serine 473 was unaffected (Fig. 1C,D). Likewise, the time-course of Akt phosphorylation upon addition of insulin for 0–15 min was unaltered by PA treatment (Fig. S1D). Myoblasts can be allowed to fuse together and become multinucleated myotubes as part of the process of muscle development. Accordingly, we investigated the effect of PA on both GLUT4 translocation and signalling to Akt in L6-GLUT4myc-hIR myotubes. As in myoblasts, PA also reduced insulin-dependent GLUT4 translocation (Fig. S2A) without affecting insulin-stimulated Akt phosphorylation (Fig. S2B). These similarities allowed us to proceed with subsequent studies employing myoblasts, which are more amenable to single-cell assays.

It has been argued that a more relevant impasse might occur at the level of AS160 (Tan et al., 2015). Given that L6-GLUT4-myc-hIR myoblasts express low levels of AS160, which prevent the reliable detection of its phosphorylation by immunoblotting (Fig. 1E), we evaluated insulin-stimulated AS160 phosphorylation in L6-GLUT4myc myoblasts stably overexpressing AS160 (L6-GLUT4myc-AS160). PA did not impair insulin-stimulated (5 nM, 15 min) AS160 phosphorylation and even potentiated this response (Fig. 1F). In parallel, we examined phosphorylation of AS160 transiently transfected into L6-GLUT4myc-hIR myoblasts. Although the phosphorylation of AS160 in these conditions did not achieve statistical significance, PA did not impair this trend at any dose tested (Fig. 1E). Hence, neither reductions in insulin-stimulated Akt nor AS160 phosphorylation are responsible for the observed impairments in insulin-stimulated GLUT4 translocation after PA treatment.

In the basal state, GLUT4 is found in both a distinct perinuclear depot and in vesicles dispersed throughout the cytosol (Jaldin-Fincati et al., 2017). The perinuclear compartment is considered to harbour the specialized GLUT4 storage vesicular compartment (GSV), which responds to insulin by propelling GLUT4-containing vesicles towards the PM (Foley and Klip, 2014). We therefore tested whether PA affects intracellular GLUT4 localization, in comparison to BSA treatment. Three main patterns of perinuclear GLUT4 localization were observed in the basal (unstimulated) state and are catalogued as follows: tight (dense depot of GLUT4 within a small radius around the nucleus), expanded (a more expanded distribution of GLUT4 relative to the nucleus), and absent (no perinuclear enrichment of GLUT4) (Fig. 2A). By visual binning, GLUT4 localized almost entirely in ‘tight’ GLUT4 compartments in BSA-treated myoblasts with very few cells displaying the ‘expanded’ or ‘absent’ phenotype (Fig. 2A,B). In contrast, PA caused a centric dispersion the GLUT4 compartment, resulting in the majority of PA-treated cells having the ‘expanded’ phenotype (Fig. 2B). A similar, small proportion of cells had the ‘absent’ phenotype in BSA- and PA-treated conditions (Fig. 2B). The intensity of GLUT4 fluorescence was also quantified along a line drawn beginning at the nuclear edge, bisecting the centre of the perinuclear compartment and projecting linearly towards the cell periphery (Fig. 2C). In BSA-treated myoblasts, intracellular GLUT4 fluorescence peaked near the nucleus and steadily decayed towards the cell periphery (Fig. 2C). In PA-treated myoblasts, peak GLUT4 fluorescence significantly shifted away from the nucleus (compared to BSA control, Table S1) and then gradually decayed towards the cell periphery (Fig. 2C). The perinuclear GLUT4 distribution in muscle and adipose cells is sensitive to insulin and re-localizes from a polar ‘conical’ to a concentric distribution (as previously reported; Dugani et al., 2008). This occurs concomitant with the migration of a cohort of GLUT4 vesicles along the cytoplasm to the PM. Given that PA vastly reduced the gain in insulin-dependent surface GLUT4, we evaluated whether the expanded perinuclear GLUT4 distribution caused by PA would nonetheless be insulin responsive. However, upon insulin stimulation, the GLUT4 compartment in PA-treated myoblasts remained expanded (Fig. S3), evincing a correlation between the expanded perinuclear GLUT4 distribution and loss of GLUT4 translocation to the membrane. We surmise that PA alters the localization of the GSV pool that is responsible for dispatching GLUT4-containing vesicles the PM upon insulin stimulation. Importantly, this alteration occurs in the basal state, prior to stimulation with insulin.

We next considered whether mis-localization of the perinuclear GLUT4 could be due to a general disruption of the Golgi or trans-Golgi network (TGN). In unstimulated cultured 3T3-L1 adipocytes (Brumfield et al., 2021), human skeletal muscle cells (Camus et al., 2020) and muscle biopsies (Knudsen et al., 2020), GLUT4 partially colocalizes with the Golgi/trans-Golgi markers GM130 and TGN46 (also known as GOLGA2 and TGOLN2, respectively). In BSA-treated L6-GLUT4myc-hIR myoblasts, GM130 localized to a tight concentric ring around the nucleus, whereas TGN46 had a more diffuse, albeit still broadly perinuclear, localization (Fig. 2D,E). Unlike GLUT4, the localization of either GM130 or TGN46 was not altered by PA, highlighting that the change in localization of GLUT4 vesicles observed upon PA-treatment is selective (Fig. 2D,E). PA did not significantly alter the overall colocalization of GLUT4 with either GM130 or TGN46, ostensibly because their colocalization is low in the first place (Manders' coefficient of <0.3).

To determine whether PA altered retrograde traffic of GLUT4, we measured the colocalization of GLUT4 with TGN46 as GLUT4 internalized from the plasma membrane for 60 min. At 30 min, significant increases in GLUT4 colocalization with TGN46 was observed in both BSA- and PA-treated cells with no significant differences between them (Fig. 2F). These observations suggest that the perinuclear expansion pertains to the subset of GLUT4 vesicles that is distinct from those stored in Golgi or trans-Golgi loci.

As GLUT4 is internalized from the PM, it diverges from other recycling cargo and moves from endosomes to a subdomain of the TGN (Foley and Klip, 2014). In L6 skeletal muscle cells, internalized GLUT4 accumulates in a syntaxin-6-(STX6)-positive subcompartment from which insulin-responsive vesicles are derived (Foley and Klip, 2014). Although STX6 is not required for the retrieval of surface GLUT4 to the perinuclear region, it is required for its insulin-responsive exocytosis (Foley and Klip, 2014). In BSA-treated cells, STX6 localized primarily to compact, crown-like structures that tended to collect on one side of the nucleus (Fig. 3A). In PA-treated cells, the STX6-positive subcompartment was dispersed, revealing a less restricted perinuclear localization that appeared more diffuse (Fig. 3A). Importantly, while PA treatment led to dispersion of the STX6-positive subcompartment, the colocalization of STX6 and GLUT4 was unchanged by PA (Fig. 3B). Line profiling across the cytoplasm quantitatively established the enrichment of STX6 fluorescence in the perinuclear region in BSA-treated myoblasts that is lost in PA-treated myoblasts due to the dispersal of the STX6-positive subcompartment (Fig. 3C; Table S2). Therefore, PA jointly alters the intracellular localization of both GLUT4 and STX6, causing dispersion of their normally compact perinuclear depots.

The intracellular GLUT4 depot is maintained by constitutive recycling between intracellular loci and the PM (reviewed in Jaldin-Fincati et al., 2017), and the maintenance of this dynamic equilibrium is critical for GLUT4 translocation. Given that PA reduced the recruitment of GLUT4 from the storage compartment to the PM in response to insulin, and the steady-state distribution of GLUT4 and STX6 was altered, we hypothesized that PA might impinge on the intracellular sorting of GLUT4 in the opposing direction – from the PM back to the STX6-positive storage compartment. To investigate this possibility, surface-facing GLUT4 was pulse-labelled with anti-Myc antibody (at 4°C to prevent internalization) followed by rewarming to allow its internalization in live cells over time. In cells that were not pre-warmed (time 0), surface GLUT4 appeared punctate and evenly distributed along the surface of the cell in both BSA- and PA-treated conditions (Fig. 3D). In BSA-treated myoblasts, surface-labelled GLUT4 began to collect at the perinuclear compartment by 20 min of rewarming, showing an increased accumulation at 30 min (Fig. 3D). In contrast, GLUT4 did not particularly relocate to the perinuclear region in PA-treated cells and instead presented a dispersed distribution at both 20 and 30 min (Fig. 3D). GLUT4 retrieved from the PM in PA-treated cells failed to localize in the perinuclear region at all, remaining dispersed throughout the cytosol at all timepoints tested (up to 1 h). Sequential Z-plane images confirm that the failure of PA-treated cells to concentrate internalizing GLUT4 in the perinuclear compartment is not an overt defect of internalization, as GLUT4 fluorescence is detected in Z-planes located in the middle of the cells (Fig. 3E). Therefore, PA impairs the sorting of externalized GLUT4 back to the perinuclear compartment and might contribute to the expansion of the steady-state expanded compartment observed in PA-treated cells (Fig. 3D).

Whereas PA treatment did not affect Akt phosphorylation, GLUT4 delivery to the PM requires the parallel activation of two signalling pathways downstream of PI3K, respectively leading to Akt and to Rac1 (Jaldin-Fincati et al., 2017). In skeletal muscle myoblasts, insulin signalling leads to dynamic cortical actin ruffling through the concerted activation of Rac1 and its effectors coflin-1, Arp2/3 and PAK proteins (Chiu et al., 2010; JeBailey et al., 2007; Tunduguru et al., 2014). This remodelled cortical actin network functions as a scaffold to provide context for efficient GLUT4 delivery to the PM (Tong et al., 2001). Previous work by our group and others has demonstrated the requirement for Rac1 in GLUT4 translocation in mature skeletal muscle (Raun et al., 2018), and for both GLUT4 translocation and actin reorganization in L6 myoblasts (Sylow et al., 2014).

To examine whether PA impaired actin remodelling, we visualized actin in live cells transfected with F-tractin conjugated to tdTomato to fluorescently label actin filaments. Beginning at 2 min and persisting until at least 6 min of insulin stimulation, dynamic cortical actin remodelling was clearly observed in BSA-treated myoblasts, which formed ruffles, but not in PA-treated myoblasts (Fig. 4A). Instead, PA-treated myoblasts showed a stable F-actin architecture that could not mount any cytoskeletal remodelling (Fig. 4A). To quantify this phenotype, we measured the ruffling index (number of actin ruffles per cell at the time of imaging, as detailed in the Materials and Methods) of BSA- and PA-treated myoblasts in response to insulin in fixed cells using phalloidin staining (Fig. 4B,C). A reviewer who was not aware of the experimental conditions quantified the ruffling index in 20–40 cells from three independent experiments. Insulin triggered the formation of an average of 3.53±0.06 (mean±s.e.m.) ruffles per cell in BSA-treated myoblasts compared to 0.58±0.22 in basal conditions. In contrast, insulin failed to induce comparable ruffling in PA-treated cells (0.17±0.05 and 0.74±0.32, for basal and insulin-stimulated cells, respectively) (Fig. 4C). Like myoblasts, actin remodelling in PA-treated myotubes also failed (Fig. S2C). Given that PA-treated cells failed to undergo the large actin remodelling events that underlie ruffling, we wondered whether the fatty acid treatment altered overall actin architecture. Interestingly, although phalloidin staining did not reveal gross differences in F-actin organization in PA-treated myoblasts in the basal state (Fig. 4B), atomic force microscopy (AFM) experiments revealed that PA significantly elevated the surface elastic modulus compared to that seen with BSA (Fig. 4D; Fig. S4). Given that the abundance and arrangement of cytoskeletal filaments are the major determinant of cellular stiffness (Pegoraro et al., 2017), these experiments argue that PA treatment alters the rigidity of the cortical actin cytoskeleton. These changes occur in the basal state, potentially precluding the subsequent response to insulin of actin filaments.

The observation that insulin failed to promote actin remodelling in PA-treated myoblasts prompted us to investigate whether Rac1 activation was altered by PA, given that Rac1 is a chief regulator of insulin-stimulated actin remodelling (Chiu et al., 2011). Downstream of insulin receptor activation, Rac1 is activated (GTP-loaded) and signals to downstream effectors (PAK1, cofilin-2 and Arp2/3) that execute actin remodelling (illustrated in Fig. 5A). Insulin promoted a 1.4-fold increase in GTP-loaded Rac1 in BSA-treated cells (Fig. 5B). Intriguingly, PA treatment elevated the basal level of Rac1 activation by 1.5-fold, similar to insulin-stimulated levels in BSA-treated cells, without altering total Rac1 expression (Fig. 5C,D). Importantly, insulin failed to elicit any further activation of Rac1 in PA-treated cells (Fig. 5B), evincing failure of the Rac1 response to insulin stimulation. It is conceivable that the elevated activation of Rac1 in the PA-treated state is an adaptive response aiming to counteract the loss of signalling and action downstream to the GTPase. Consistent with the arrest of actin remodelling in PA-treated myoblasts, transient overexpression of constitutively active Rac1 (CA-Rac1) failed to induce actin ruffles in PA-treated myoblasts, unlike the ruffling response observed in BSA-treated cells (Fig. S5). In line with this observation, CA-Rac1 overexpression could not restore GLUT4 translocation in PA-treated cells (Fig. S5), supporting the notion that the defect in insulin-stimulated actin remodelling might arise downstream of Rac1.

Productive actin remodelling is a cyclic process that requires both actin polymerization and de-polymerization, and the p21-activated kinases (PAKs) are Rac1 effectors that regulate both processes (Tsakiridis et al., 1996; Tunduguru et al., 2014). In response to insulin stimulation, Rac1 binding to a PAK protein disassembles the auto-inhibitory homodimer conformation and leads to PAK autophosphorylation and activation. PAK1 is required for glucose uptake in skeletal muscle cells (Tunduguru et al., 2014) and PAK2 is required in skeletal muscle tissue (Møller et al., 2020). Insulin evoked phosphorylation of both PAK1 and PAK2 in L6-GLUT4myc-hIR myoblasts and PA inhibited the hormonal response of both isoforms (Fig. 5C,E,F).

Activated PAK contributes to actin depolymerization via dephosphorylation and activation of cofilin (Tunduguru et al., 2014). Cofilin is an actin-binding protein that increases actin-filament turnover through its actin-severing activity (Maciver and Hussey, 2002) thereby elevating the rate of monomer liberation from the pointed end of actin filaments. Cofilin phosphorylation at serine 3 prevents its binding to actin (Maciver and Hussey, 2002). In the basal state, serine 3 was phosphorylated, and insulin caused its de-phosphorylation (via the phosphatase slingshot), liberating monomers for iterative actin cycling (Chiu et al., 2010). PA reduced insulin-dependent cofilin dephosphorylation in a dose-dependent fashion (Fig. 5C,G). Taken together, defective insulin signalling downstream of Rac1 impairs activation of PAK proteins and, subsequently of cofilin-2, resulting in a local net loss of free actin monomers available for branched filament formation. This scenario likely accounts for the failure in cortical actin remodelling in PA-treated myoblasts.

During successful actin remodelling cycles, actin monomers liberated by actin-severing activity of cofilin-2 are in turn added by the Arp2/3 complex to generate branches on existing filaments (Jaldin-Fincati et al., 2017). Accordingly, we investigated the behaviour of ArpC2 in PA-treated myoblasts, as this is one of the seven subunits of the Arp2/3 complex previously shown to be required for insulin-stimulated GLUT4 translocation and actin remodelling (Chiu et al., 2010). In unstimulated BSA-treated cells, ArpC2 localized to both the cytosol and peripheral cell borders, becoming further enriched in peripheral membrane regions of active actin remodelling upon insulin stimulation (Fig. 5H). In contrast, ArpC2 was not detected at peripheral cell borders in unstimulated PA-treated cells and was instead visualized in the cytosol with dense accumulation in the perinuclear region of the myoblasts (Fig. 5H). This was not caused by any changes in the total ArpC2 protein content (Fig. 5I). Insulin did not evoke any changes in ArpC2 localization in PA-treated cells (Fig. 5H). Notably, PA brought about persistent changes in the localization of ArpC2 in the basal state, which might impinge on its subsequent ability to respond to insulin.

Collectively, these findings indicate that PA impairs insulin-dependent actin reorganization in two ways: by increasing cortical actin stiffness and by precluding Rac1 signalling to its downstream effectors involved in both actin polymerization and de-polymerization, responses that are established to be critical elements for productive GLUT4 translocation.

To investigate whether the effects of PA could be reversed, myoblasts treated with or without BSA or PA 18 h, were switched for 24 h to growth medium without added lipids, before stimulation with insulin. The selection of a 24 h withdrawal period was informed by experiments demonstrating that PA-induced insulin resistance does not occur after short exposures to PA (0–8 h; Fig. S6); hence, short periods of PA discontinuation might be insufficient to permit reversal of these effects. The 24 h withdrawal from PA reverted GLUT4 localization to a tight, perinuclear depot (Fig. 6A), enabled insulin-stimulated actin remodelling (Fig. 6A) and restored GLUT4 translocation (Fig. 6B). Therefore, although PA impairs GLUT4 localization, actin dynamics and GLUT4 translocation, these effects are not irreversible. These findings allowed us next to examine whether targeting selective mechanisms that have been postulated to cause insulin resistance could restore insulin responsiveness in PA-treated myoblasts.

A variety of mechanisms have been heralded to contribute to insulin resistance caused by saturated fatty acids, such as ceramide accumulation, oxidative stress, palmitoylation and ER stress (reviewed in Martins et al., 2012). Although a combination of mechanisms undoubtedly contributes to the development of insulin resistance in vivo, the cell culture system in this study permits the investigation of cell-autonomous causes of skeletal muscle insulin resistance. Indeed, although PA elevates intracellular ceramides, they are not responsible for cell-autonomous insulin resistance in skeletal muscle myotubes (Pillon et al., 2018), despite being widely implicated in the development of insulin resistance in vivo (Chaurasia and Summers, 2021).

We therefore sought to determine whether the defects in GLUT4 translocation, GLUT4 perinuclear localization or actin remodelling provoked by PA treatment could be restored by mitigating oxidative stress, palmitoylation or ER stress. To this end, the incubation with either BSA or PA additionally included the antioxidant N-acetyl cysteine (NAC) (Pedre et al., 2021), the competitive palmitoylation inhibitor (2-bromopalmitate; 2-BP) (Davda et al., 2013) or the chemical chaperone 4-phenylbutyric acid (4-PBA) (Kolb et al., 2015). Given that PA impinges on GLUT4 translocation and at least two core nodes required for this process (proper intracellular GLUT4 sorting and actin remodelling), we investigated the effects of the compounds on each node to determine whether we could identify a single mechanism for the dysregulation of each node. Recent work implicates the mitochondrial oxidative burden in the development of lipid-induced muscle insulin resistance (Fiorenza et al., 2023 preprint). We investigated the effect of total cellular oxidative stress on insulin resistance using NAC. Treatment with NAC did not restore either actin remodelling or GLUT4 localization in PA-treated cells, suggesting t­hat oxidative stress does not contribute to PA-induced insulin resistance in this context (Fig. S7). However, treatment with 2-BP restored GLUT4 localization (to a tight, perinuclear depot not seen in PA-treated cells), but not actin remodelling, in PA-treated cells (Fig. 7A). Conversely, treatment with 4-PBA restored actin remodelling, but not GLUT4 localization in PA-treated cells (Fig. 7B). Therefore, palmitoylation contributes to PA-induced mislocalization of GLUT4, whereas ER stress contributes to insulin resistance of actin remodelling. Neither NAC (Fig. S7), 2-BP (Fig. 7A) nor 4-PBA (Fig. 7B) were able to preserve insulin-stimulated GLUT4 translocation in PA-treated myoblasts. Hence, mitigating a single mechanism of PA-elicited insulin-resistance is insufficient to preserve GLUT4 translocation.

Given that different mechanisms appeared to contribute to the altered sorting of GLUT4 and to insulin resistance of actin remodelling, we hypothesized that joint mitigation of ER stress and inhibition of palmitoylation might achieve preservation of GLUT4 translocation in PA-treated cells. However, insulin-induced GLUT4 translocation remained inhibited in cells co-treated with 2-BP, 4-PBA and PA (Fig. 7C). Therefore, PA-induced insulin resistance of GLUT4 is a multi-hit phenomenon and includes at least these two nodes and other so-far unidentified inputs additional to ER stress and protein palmitoylation. These results reveal that cell-autonomous insulin resistance is multi-factorial, adding to the level of complexity of the metabolic dysregulation that occurs in insulin resistance in vivo, a complex trait involving multiple organs, pathways and mechanisms.

The efficacy of unsaturated fatty acids in mitigating PA-induced insulin resistance of glucose uptake in skeletal muscle cells is well documented (Dimopoulos et al., 2006; Sawada et al., 2012). However, whether unsaturated fatty acids restore insulin-stimulated actin remodelling and intracellular GLUT4 localization in PA-treated cells was unknown. As palmitoleate (PO, C:18:1) is the monounsaturated counterpart to PA, we examined whether co-treatment with PO might preserve insulin sensitivity in the presence of PA. To this end, we treated cells with either BSA, PA or PO, or co-treated cells with PA and PO (0.350 mM each). GLUT4 translocation, actin remodelling and GLUT4 localization in PO-treated cells were equivalent to those in control BSA-treated cells, indicating that unlike its saturated counterpart, PO does not cause insulin resistance (Fig. S8). In contrast to the deleterious effects of PA, co-treatment with PA and PO preserved GLUT4 localization (Fig. 8A), insulin-stimulated actin remodelling (Fig. 8A) and insulin-dependent GLUT4 translocation (Fig. 8B). Therefore, exposing myoblasts to PA in the presence of PO preserves insulin responsiveness, and might become a useful tool in dissecting the full complement of signals and mechanisms whereby saturated fats confer insulin resistance to GLUT4.

These results emphasize that, despite the multipronged action of PA, its consequent impairment in GLUT4 translocation was not irreparable. That co-treatment with PO fully restored GLUT4 translocation is consistent with previous work showing the protective effects of unsaturated fats on saturated fat-induced insulin resistance (Gao et al., 2009). It further extends the actions of PO to saturated fat-induced defects in actin remodelling and perinuclear GLUT4 localization. The beneficial action of PO might result from metabolic actions precluding the PA-induced negative actions, or it might be restorative.

Skeletal muscle is a critical regulator of whole-body glucose homeostasis, and it is vulnerable to developing insulin resistance following sustained nutrient excess as is often seen in the Western diet, particularly with excess of saturated fats. A defining feature of insulin resistance is the failure to mount an appropriate, surface translocation of GLUT4 in response to insulin. Strikingly, there are no therapies that directly target GLUT4 translocation and thus, there is pressing need to determine precisely how, why and when GLUT4 translocation fails. Here, we show that the saturated fat PA, but not its unsaturated counterpart PO, causes a cell-autonomous reduction of insulin-stimulated GLUT4 translocation in cultured myoblasts and myotubes without impairing insulin signalling to Akt and AS160, highlighting that the defect must occur independently of these signals. Instead, we find that even in the absence of insulin, PA alters two fundamental mechanical aspects that are critical for subsequent insulin response – the intracellular sorting of GLUT4 and cortical F-actin architecture. These underlying alterations limit the ability of myocytes to undergo the appropriate actin remodelling that allows for productive GLUT4 translocation and insertion into the plasma membrane upon insulin exposure. Although the failure of insulin-dependent actin remodelling was prevented by mitigating ER stress with 4-PBA, intracellular mislocalization of GLUT4 persisted. By contrast, the altered GLUT4 sorting was prevented by inhibiting palmitoylation with 2-BP, yet insulin-dependent actin remodelling was still impaired. Hence, defects in insulin-induced GLUT4 translocation and cellular ruffling following PA treatment are conferred by unique alterations in the basal state. This dissociates the altered intracellular distribution of GLUT4 from insulin signalling, confirmed by the lack of inhibition of Akt or AS160 in response to the hormone (Fig. 1).

In the basal state, GLUT4 undergoes dynamic recycling to and from the PM, transitioning through recycling endosomes, and some of the transporters reach and repopulate the GSVs with the help of STX6 (Jaldin-Fincati et al., 2017). Previous work from our group has demonstrated that C2-ceramide disperses this subcompartment leading to impaired insulin-stimulated GLUT4 exocytosis (Foley and Klip, 2014). Like C2-ceramide, PA caused expansion of the normally compact, perinuclear depot of GLUT4 in the basal state. Although these GLUT4 vesicles remained populated by STX6 in the steady state, their spatial distribution was significantly altered. Indeed, a couple of studies have reported altered intracellular distribution of GLUT4 in subcellular fractions isolated from insulin-resistant muscles under basal conditions (Garvey et al., 1998; Klip et al., 1990), heralding the possibility that defects in insulin-stimulated GLUT4 translocation might originate in intracellular sorting of the transporter. Importantly, total GLUT4 levels were not affected in those studies or the current one, highlighting that insulin resistance is not simply a consequence of loss of transporters, and rather that it is the result of their altered sorting that precludes them reaching the PM upon insulin stimulation.

Additionally, it is possible that PA misplaces GLUT4 such that it cannot be accessed by later downstream effectors responsible for vesicle mobilization (i.e. insulin-responsive Rab8A and Rab13; reviewed in Jaldin-Fincati et al., 2017). Whether PA impairs only the mobilization of insulin-dependent cargoes or exerts a broader alteration in cargo mobilization regardless of stimulus could be investigated as part of future studies. However, our experiments show that neither the Golgi nor the TGN at large were mislocalized in a manner similar to GLUT4 in PA-treated myoblasts. Similarly, in a preliminary experiment, we did not observe mislocalization of the v-SNARE VAMP2 in response to PA, suggesting that GLUT4 vesicles are mis-sorted before acquiring this v-SNARE, which defines insulin-responsive vesicles (V.L.T., unpublished observations).

Prior to insulin stimulation, cellular stiffness was significantly higher in PA-treated than in BSA-treated cells, as measured by AFM. Because the abundance and arrangement of cytoskeletal filaments is a major determinant of cellular stiffness (Pegoraro et al., 2017), implying that cytoskeletal rearrangements presumably drive the observed changes in cellular rigidity. Indeed, both insulin resistance and PA have been shown to alter basal-state F-actin in skeletal muscle tissue (Grice et al., 2019) and myotubes (Habegger et al., 2012), respectively. Although we did not observe a comparable global loss of F-actin, it is clear that the actin cytoskeleton is affected in insulin resistance, and moreover, that a dynamic actin cytoskeleton is important for insulin-stimulated GLUT4 translocation (Tong et al., 2001; Török et al., 2004). That PA alters the actin cytoskeleton is not specific to muscle cells, as it also occurs in human dermal lymphatic endothelial cells (Tokarz et al., 2023) and podocytes (Xu et al., 2015).

Consistent with increased cell stiffness, PA increased basal activation of Rac1. This Rho-family GTPase promotes polymerization to form branched F-actin, which has greater load-bearing capacity than linear filaments (Bieling et al., 2016). Indeed, Rac1 activation increases cell stiffness in a variety of cell types (Garitano-Trojaola et al., 2021; Kunschmann et al., 2019). Additionally, PA might affect the architecture of the actin cytoskeleton by regulating the ability of actin filaments to assemble into higher-order structures (such as formation of crosslinks, bundles and networks), which is a compelling possibility to consider. Insulin-dependent actin remodelling events are believed to permit and regulate the physical movement of GLUT4 vesicles for productive fusion with the PM (Lopez et al., 2009; Tong et al., 2001). A rigid cortical actin architecture in PA-treated cells might represent a significant physical barrier for GLUT4 vesicle insertion into the PM, a step that is defective in insulin-resistant adipocytes (Lizunov et al., 2013).

These findings suggest that PA causes priming deficiencies at the level of the basal cellular architecture and processes that might confer resistance upon subsequent stimulation with insulin. Although most studies focus on post-insulin receptor signalling events, successful GLUT4 translocation requires that GLUT4 vesicles are both available and capable of successful translocation. In this context, insulin responsiveness is determined by events independent of insulin binding to its receptor, requiring that all of the physical signalling proteins are ready and able to respond to the hormone. Our findings in response to a saturated fat resonate with the recently reported alterations in the basal-state proteome of induced pluripotent stem cells (iPSCs) derived from individuals with type 2 diabetes (Batista et al., 2020).

A large proportion of the existing studies on PA-induced insulin resistance in skeletal muscle focus on a single mechanism [i.e. ceramide synthesis (Pillon et al., 2018), oxidative stress (Yuzefovych et al., 2010), ER stress (Panzhinskiy et al., 2013), inflammation (Coll et al., 2008), and others (reviewed in Sears and Perry, 2015)] and do not investigate the cause of insulin resistance at the resolution of individual steps required for GLUT4 translocation. As we identified two distinct nodes of PA-induced insulin resistance (GLUT4 mislocalization and cytoskeletal stiffness precluding actin remodelling), we endeavoured to determine whether they arose through a common mechanism. This was not the case, as GLUT4 localization was restored by inhibition of palmitoylation with 2-BP, whereas actin remodelling was restored by mitigation of ER stress with the chemical chaperone 4-PBA. Habegger et al. (2012) have successfully restored GLUT4 translocation in myotubes upon cholesterol removal with β-cyclodextrin; however, in preliminary experiments, this treatment was not equally successful in hIR-GLUT4myc myoblasts (V.L.T., unpublished observations).

Our findings show that PA exerts pleiotropic effects on the GLUT4 translocation process, impinging on multiple points throughout the pathway via distinct mechanisms in a single system. That actin remodelling (but not GLUT4 localization) was preserved in PA-treated cells upon co-treatment with the chemical chaperone 4-PBA is consistent with the previous observation that actin bundling is dysregulated by ER stress (Borradaile et al., 2006) and that ER stress alters actin turnover dynamics and intracellular filament organization (Bedard et al., 2019). Unlike actin remodelling, GLUT4 localization was not preserved upon treatment with 4-PBA. Instead, inhibition of palmitoylation with 2-BP preserved the compact, perinuclear localization of GLUT4 seen in BSA-treated cells. Protein palmitoylation plays a critical role in localizing proteins to functional membrane subdomains (Guan and Fierke, 2011). Indeed, GLUT4 as well as many key constituents of GSVs [such as STX6, VAMP2 and IRAP (also known as LNPEP)] are palmitoylated (Ren et al., 2013).

In conclusion, we propose that PA causes the insulin resistance phenotype of diminished insulin-stimulated GLUT4 translocation, along with altering intracellular GLUT4 localization and actin cytoskeletal stiffness, which impairs actin remodelling in response to Rac1, without any modification in proximal insulin receptor signalling. These mechanisms likely contribute to the diminished GLUT4 translocation in response to insulin, but certainly additional mechanisms are required and remain to be identified. The two defects described originate in the basal state, possibly ‘priming’ cells for insulin resistance upon later stimulation with the hormone. Inhibiting protein palmitoylation and mitigating ER stress and inhibiting palmitoylation restores GLUT4 perinuclear sorting and actin remodelling, respectively, but neither restored GLUT4 translocation. Such multi-modal inputs suggest that mitigation of a single cause is insufficient to restore the integrated process of GLUT4 translocation. Insulin resistance is specific to the unsaturated fat PA, but not the monounsaturated fat PO, and full insulin responsiveness can be preserved by co-treatment with the latter. These findings are important for our understanding of how, when and why GLUT4 translocation fails during insulin resistance caused by nutrient excess and might be important for the future development of therapies directly targeting GLUT4 translocation.

Low-endotoxin BSA (A8806) and PA (P9767) were from Sigma-Aldrich (St Louis, MO, USA); PO (10009871) was from Cayman Chemicals (Ann Arbor, MI, USA. 2-BP (238422); 4-PBA (SML0309), latrunculin A (428021) and NAC (A7250) were from Sigma-Aldrich. Antibodies used for immunoblotting were against α-actinin (Millipore Sigma, A7811), β-actin (Cell Signaling, 4790), p-Akt S473 (Cell Signaling, 4060), p-Akt T308 (Cell Signaling, 9275), Akt (Cell Signaling, 2920), ArpC2 (Millipore, 07-227), p-AS160 T642 (Cell Signaling, 4288), AS160 (Cell Signaling, 2670), p-cofilin S3 (Cell Signaling, 3311), cofilin (Cell Signaling, 3312), Myc (Santa Cruz Biotechnology, 9E10), p-PAK1/2 Thr423/402 (Cell Signaling, 2601), PAK1/2 (Cell Signaling, 2604) and Rac1 (Abcam, 33186), and IRDye 800CW goat anti-rabbit IgG (LiCor, 9253211) and IRDye 680LT goat anti-mouse IgG (92668070). Antibodies and reagents used for immunofluorescence were: antibodies against ArpC2 (Millipore, 07-227) and STX6 (BD Biosciences, 610636); phalloidin (conjugated to Alexa Fluor 555) (Thermo Fisher Scientific, A34055); goat-anti rabbit IgG secondary antibody (Alexa Fluor 488) (Thermo Fisher Scientific, A11008) and goat-anti mouse IgG secondary antibody (Alexa Fluor 647) (Thermo Fisher Scientific, A21235); and Dako fluorescence mounting medium (S3023) from Aligent. The plasmid encoding CA-Rac1-GFP was kindly provided by Dr Sergio Grinstein (Hospital for Sick Children, Toronto, Ontario, Canada).

Cells were derived from lab stocks (available for purchase via https://www.kerafast.com/item/840/l6-glut4myc-rat-myoblast-cell-line). Generation of the cell lines stably expressing human insulin receptor (hIR) or AS160 was performed in our laboratory. L6-GLUT4myc-hIR and L6-GLUT4myc-AS160 were cultured in α-MEM (StemCell Technologies) containing 10% fetal bovine serum and antibiotics at 37°C with 5% CO₂. Prior to insulin stimulation, cells were incubated for 3 h in serum-free α-MEM. Myoblasts were differentiated into myotubes by reducing the culture medium fetal bovine concentration to 2% for 5 days after myoblasts reached confluence.

Solutions of PA or PO were prepared in 50% ethanol and heated at 50°C to generate 200 mM stocks. These stocks were diluted in 10.5% BSA solution (low-endotoxin BSA dissolved in serum-free medium) 25× and conjugated with agitation at 40°C for 2 h to generate 8 mM stock solutions. These solutions were stored in aliquots at −20C and thawed upon use. Incubation with lipids was performed for the time indicated.

L6-GLUT4myc lysates were collected in RIPA buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% SDS) supplemented with 5 mM NaF, 1 mM EDTA, 1 mM PMSF, 1 mM Na₃VO₄ and protease inhibitors (Sigma Aldrich, P3840). The Pierce BCA protein assay kit (Thermo Fisher Scientific, 23225) was used to determine protein concentration. 20–40 µg of total protein was diluted in Laemmli buffer with dithiothreitol (DTT) and separated by SDS-PAGE, transferred to nitrocellulose membrane, blocked with 5% BSA and incubated overnight with primary antibodies [1:1000 dilutions, prepared in TBS-tween (Tris-buffer saline with 0.1% Tween 20)] at 4°C. Fluorescent secondary antibodies from LI-COR were incubated at room temperature for 45 min (1:10,000 dilutions, prepared in TBS-tween) protected from light. The membranes were imaged using the Odyssey Fc Imager (LI-COR) and quantified using Image Studio 4.0 software (LI-COR). Full, uncropped blot images are available in the supplementary materials. Full uncropped images of blots in this study are shown in Fig. S9.

L6-GLUT4myc-hIR cells were grown to the desired confluence, treated with or without lipids and chemical inhibitors as described, then fixed in 4% paraformaldehyde (PFA) for 10 min. To detect total proteins, permeabilization was accomplished with 0.25% Triton X-100 for 10 min, followed by blocking in 2% BSA for 30 min. To detect surface proteins, cells were not permeabilized and blocked immediately after fixation in 2% BSA for 30 min. Primary antibodies were always diluted in 0.2% BSA, used at 1:100 dilutions in 2% BSA-PBS, and incubated for 45 min–1 h at 37°C. Cells were incubated with secondary antibodies (1:1000), DAPI (1:1000) and Alexa Fluor 555 phalloidin (where indicated, 1:200) for 45 min at 37°C. Coverslips were mounted on microscope slides with Dako mounting medium. An Olympus Quorum spinning disc confocal microscope (60× NA 1.35 objective) was used to capture images processed by Volocity 6.1.2 software (Perkin Elmer). Where indicated, quantification of fluorescence intensity was performed using ImageJ.

L6-GLUT4myc-hIR myoblasts were grown to confluence and serum starved for 2–3 h prior to insulin stimulation. Myoblasts were stimulated with 0.1 nM insulin (Eli Lilly, Indianapolis, USA) for 15 min at 37°C. Cells were placed on ice after stimulation, and immediately washed twice with ice-cold PBS. All steps thereafter were performed on ice. Cells were fixed in 4% PFA for 20 min, quenched with 0.1 M glycine for 20 min and blocked in 5% goat serum for 30 min. Cells were incubated with primary antibody against the exofacial Myc epitope of GLUT4 (1:500, anti-c-Myc, Sigma, C4956) for 1 h. Cells were extensively washed with PBS prior to incubation with HRP-conjugated goat anti-rabbit IgG (1:1000) for 1 h. After extensive PBS washing, incubation with OPD reagent produced a colorimetric reaction that was stopped after 3.5 min with the addition of 3 M HCl. Optical absorbance was measured at 492 nm.

GTP-bound Rac1 was detected using the Rac1 G-LISA assay from Cytoskeleton (BK128) according to the manufacturer's instructions. Briefly, cells were serum starved and stimulated with insulin for 2 min. Cells were rapidly lysed in G-LISA Lysis Buffer (Cytoskeleton) and flash frozen in liquid nitrogen. Protein concentration between lysates was equalized prior to the beginning of the experimental G-LISA protocol. Activated (GTP-bound) Rac1 was immobilized to the plate and incubated with primary antibody against Rac1 for 45 min at room temperature, followed by a 45-min incubation with HRP-conjugated secondary antibody. The colorimetric reaction produced during detection was read by spectrophotometer at 492 nm and data was analysed according to manufacturer's instructions.

After a 2 h serum starvation, cells were placed on ice to arrest intracellular traffic. Cells were then washed in PBS and blocked with 5% goat serum on ice for 20 min. Cells were incubated with anti-Myc antibody (1:100, Sigma, C4956) for 1 h at 4°C to label surface GLUT4. After incubation with anti-Myc antibody, cells were washed in PBS to remove any unbound antibody. To allow for internalization of surface-labelled GLUT4, cells were re-warmed for the indicated timepoints in serum-free medium at 37°C. The experiment was stopped by placing cells in 4% PFA at 4°C for 20 min. Cells were processed according to the immunofluorescence protocol above.

The elastic modulus of myoblasts was measured by atomic force microscopy using a Zeiss Axiovert 200 M inverted epifluorescence microscope with a Nanowizard 4 (JPK Instruments). Cells were cultured on glass coverslips and maintained in serum free medium at 37°C, 5% CO₂ for the duration of imaging. Indentations were made using a gold cantilever with a polystyrene bead (10 µm). The approach rate was 2 µm/s. Force–indentation profiles were generated on a 20×20 µm area to visualize surface topography. Force–indentation curves were then obtained from multiple points on the cell surface (n=15–30 per cell for 18–20 cells per condition) and analysed using the Hertz cone model (Giessibl, 2003) to determine the elastic modulus for each cell.

The fluorescence intensity of pixels from the channel corresponding to GLUT4 fluorescence was quantified along a straight line drawn from the centre of the perinuclear GLUT4 compartment, beginning at the nucleus and ending at the cell edge to produce a line profile. Intensity values were plotted as distance from the nucleus on the y-axis and intensity on the x-axis. Quantification was performed in ImageJ.

The ruffling index was quantified as previously described (Cox et al., 1997). Briefly, to quantify the ruffling index, an independent reviewer who was not aware of the experimental conditions scored 20–40 cells per condition from three independent experiments as follows: eligible ruffles were defined as regions of dense phalloidin staining extending into the dorsal plane when visualized by fluorescence microscopy. In cases with multiple ruffles, classification as separate or single ruffles was based on their separation in the xy plane. Total counted ruffles were normalized to number of cells to compute an average number of ruffles per cell, defined as the ‘ruffling index’ or ruffles per cell.

Manders' coefficients were calculated using Volocity 6.1.2 software (Perkin Elmer).

Analyses were performed using Prism 9.0 software (GraphPad Software). Two-way ANOVA with Tukey's multiple comparisons test was performed unless otherwise stated in figure legends. Statistical significance is as described in figure legends.

We thank Dr Philip J. Bilan for strategic guidance and valuable input throughout the study, and Zhi Liu and Humayon Akhuanzada for technical assistance. We thank Dr Kimberly Lau and Paul Paroutis for their training and assistance in use of the microscopy equipment in the Imaging Facility at The Hospital for Sick Children.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261300.reviewer-comments.pdf