Oxidative stress elicited by modifying the ceramide acyl chain length reduces the rate of clathrin-mediated endocytosis

The lipid composition of the plasma membrane plays a vital role in regulating clathrin-mediated endocytosis (CME) (Lauwers et al., 2016), and sphingolipids (SLs) are of particular importance in this regard (Cheng et al., 2006). It is perhaps not surprising that SLs play such a critical role in regulating CME since SLs have distinct biophysical properties, and together with cholesterol, ceramide and sphingomyelin (SM) increase the order and reduce the fluidity of membranes, leading to increased rigidity, tighter packing and formation of ceramide-enriched microdomains (Pinto et al., 2013). These microdomains enable amplification of signaling pathways by reorganizing the membrane and clustering specific signaling components (Zhang et al., 2009).

Most studies examining the role of SLs in CME have been performed using inhibitors of SL biosynthesis (Meyer et al., 2012) or by altering levels of specific SL classes, such as SM (Shakor et al., 2011). We previously demonstrated that altering the SL acyl chain length also leads to changes in membrane biophysical properties (Silva et al., 2012), which may be one of the reasons that ceramide synthase 2 (CerS2) null mice, which are unable to synthesize very long-chain SLs (Pewzner-Jung et al., 2010a,b), displays a variety of phenotypes related to ligand and receptor internalization (Ali et al., 2013; Park et al., 2012, 2014). However, mechanistic information on how CME is regulated by the SL acyl chain composition is unavailable. We now analyze CME in astrocytes that have been cultured from CerS2 null mice, and show that they contain elevated levels of reactive oxygen species (ROS), which trigger a reduction in the transcription of Hsc70 (also known as HSPA8) – a protein involved in uncoating clathrin vesicles – and as a consequence, reduces the rate of CME but not of endocytosis via caveolin-mediated pathways. We suggest that, at least in astrocytes, CME is not directly regulated by the SL acyl chain composition but rather via a downstream pathway that is modulated by the oxidative state of the cell.

In mammals, endocytosis of nutrients by astrocytes is important for normal neuronal function (Jiang and Chen, 2009). Astrocytes were isolated from wild-type (WT) and CerS2 null mice. The cultures contained ∼95% astrocytes and <0.1% microglia [using an antibody against glial fibrillary acidic protein (GFAP) to detect astrocytes and against CD68 to detect microglia] (Fig. 1A). As expected, in CerS2 null astrocytes, mRNA levels of CerS2 were undetectable and mRNA levels of the other five mammalian CerS proteins (Levy and Futerman, 2010) were unaltered (Fig. 1B). Likewise, CerS2 activity was undetectable in CerS2 null astrocytes using C24:1-CoA as substrate, although a small amount of residual activity could be detected using C22:0-CoA (Fig. 1C), probably due to the partial overlapping substrate specificities of CerS4 and CerS2 (Riebeling et al., 2003). Analysis of the astrocyte SL composition was consistent with mRNA and CerS-protein activity levels, in so far as C24:1-ceramide was barely detectable in CerS2 null astrocytes [as were levels of the more complex SLs, C24:1-SM and C24:1-glucosylceramide (not shown)], and C22:0-ceramide levels were reduced by ∼35% compared to those in WT astrocytes (Fig. 1D). Total ceramide levels were essentially unaltered (Fig. 1D) due to increased levels of C16:0- and C18:0-ceramide, similar to changes in the SL profile in CerS2 null mouse liver (Pewzner-Jung et al., 2010b). However, unlike in liver, levels of the long-chain bases – sphinganine (3.9±0.7 pmol/mg in WT and 5.6±1.1 pmol/mg in CerS2 null mice; mean±s.d.) and sphingosine (86.4±3.4 pmol/mg in WT and 75.2±3.3 pmol/mg in CerS2 null mice) – were essentially unaltered.

We next examined the rate of CME in astrocytes using two fluorescently labeled ligands (Keyel et al., 2006). Both 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine (Dil)-tagged low density lipoprotein (Dil–LDL) and Cy5–transferrin (Cy5–Tf) were internalized at a slower rate in CerS2 null astrocytes than in WT astrocytes (Fig. 2A,B). In contrast, the rate of caveolin-mediated endocytosis, measured using BODIPY–lactosylceramide (BODIPY–LacCer) (Singh et al., 2007), was unaltered (Fig. 2C). The differences in the kinetic parameters of uptake between Cy5–Tf (Fig. 2A) and Dil–LDL (Fig. 2B) is probably due to differences in the rate of recycling and sorting and between LDL and Tf receptors (Ghosh et al., 1994).

LDL receptor and Tf receptor levels were unchanged (Fig. 3A), suggesting that the reason for the reduced rate of CME is related to one or other aspect of CME rather than reduced levels of the receptors that bind to the ligands. Indeed, levels of the clathrin heavy chain were elevated 3.5-fold (Fig. 3A), and distinct large puncta of the clathrin heavy chain were observed in the cytoplasm of CerS2 null astrocytes (Fig. 3B), similar to that observed previously in mice that lack the clathrin-coated vesicle uncoating protein auxilin-1 (Yim et al., 2010). In the case of CerS2 null mouse astrocytes, levels of auxilin-1 mRNA were unchanged (Fig. 3C), but levels of heat shock cognate 70 (Hsc70) mRNA were reduced by ∼threefold (Fig. 3C), and protein levels were reduced by 25±5% (Fig. 3D). Hsc70 also plays a key role in the uncoating of clathrin-coated vesicles (Greene and Eisenberg, 1990).

Since Hsc70 mRNA levels were reduced (Fig. 3C), we next examined the levels of two transcription factors that regulate Hsc70 expression, namely specificity protein 1 (Sp1) (Wilke et al., 2000) and heat shock factor 1 (HSF-1) (Yasuhara et al., 2011). Levels of Sp1 were reduced by 35±4% (mean±s.d.), whereas levels of HSF-1 were unaltered (Fig. 4A); the reduction in Sp1 protein levels was not due to changes in its transcription (Fig. 4B). Two putative binding motifs for Sp1 are found in positions −174 and −68 from the transcription start site of Hsc70 (Fig. 4C). Chromatin immunoprecipitation (ChIP) demonstrated that Sp1 binds to position −68, whereas no binding of Sp1 to position −174 could be detected (Fig. 4D). Sp1 binding to position −68 was significantly reduced in CerS2 null astrocytes (Fig. 4D). Levels of two other Sp1 targets, glycolipid transfer protein (Zou et al., 2011) and fatty acid synthase (Archer, 2011) were also reduced, but levels of two predicted non-targets of Sp1, GFAP and toll-like receptor 4, were unaltered (Fig. 4E). These results suggest that Hsc70 is transcriptionally regulated by Sp1 in CerS2 null astrocytes, and that additional cellular pathways regulated by Sp1 might also be affected in CerS2 null astrocytes.

Sp1 is sensitive to its O-glycosylation (GlcNAc) state in so far as Sp1 hypo-glycosylation leads to its proteasomal degradation (Han and Kudlow, 1997). ROS have been shown previously to induce proteasome-mediated reduction of Sp1 levels through reduction of Sp1 O-glycosylation (Kim et al., 2012), whereas N-acetyl cysteine (NAC) has been shown to mitigate this effect (Hsin et al., 2010). Thus, we next measured ROS levels in CerS2 null astrocytes, which were elevated ∼fourfold (Fig. 5A), along with a 60±18% (mean±s.d.) increase in levels of 3-nitrotyrosine (Fig. 5B). ROS overproduction in CerS2 null astrocytes is likely to be due to reduced activity of mitochondrial respiratory chain complex IV (also known as COX) (Fig. 5C), similar to that observed in the liver from CerS2 null mice (Zigdon et al., 2013). Based on these results, we suggest a pathway exists by which oxidative stress in CerS2 null astrocytes results in lower Sp1 levels and hence lower levels of Hsc70, which directly affects the rate of CME. This was tested by inducing ROS, using H₂O₂, in WT astrocytes, which resulted in a slower rate of uptake of Cy5–Tf, and by ameliorating ROS levels in CerS2 null astrocytes with NAC, which resulted in increased rates of Cy5–Tf uptake (Fig. 6A,B). Similar results were seen with Dil–LDL (Fig. 6C,D). Concomitantly, H₂O₂ treatment decreased Sp1 and Hsc70 levels in WT astrocytes to levels similar to those in untreated CerS2 null astrocytes, whereas treatment with NAC increased their levels in CerS2 null astrocytes to the same levels as in untreated WT astrocytes (Fig. 6E). Neither ROS nor NAC affected LDL receptor or Tf receptor levels in WT or CerS2 null astrocytes (Fig. 6E).

In the current study, we demonstrate a new mechanism by which altering the SL acyl chain length can affect the rate of CME, which is apparently independent of changes in membrane biophysical properties. In this pathway, oxidative stress, caused by ROS overproduction via mitochondrial complex IV dysfunction, as observed in hepatocytes upon loss of very-long acyl chain ceramides and elevation of long-chain ceramides (Zigdon et al., 2013), results in downregulation of the transcription factor Sp1, which binds to the promoter region of Hsc70, a key player involved in removing the clathrin coat. As a result, clathrin aggregates in the cytoplasm, rendering it inaccessible for additional rounds of CME (Fig. 7). Disruption of another uncoating protein, auxilin-1 (Yim et al., 2010), results in a similar phenotype, as does knockdown of Hsc70 (Yu et al., 2014), whereas a complete knockout of Hsc70 in mice is embryonically lethal (Florin et al., 2004), and auxilin-1 knockout mice exhibit elevated levels of clathrin heavy chain (Hirst et al., 2008) and reduced CME (Yim et al., 2010).

SLs can reside in lipid rafts (Simons and van Meer, 1988), platforms through which caveolae are formed and, thereby, they mediate uptake of ligands through caveolin-mediated endocytosis (Mayor and Pagano, 2007). Although we observed significant changes in the SL profile of CerS2 null astrocytes, caveolin-mediated endocytosis was unaltered, presumably since total ceramide levels were unaltered in CerS2 null astrocytes.

We have previously observed a number of changes in receptor activation and/or internalization in hepatocytes. For instance, altering the SL acyl chain composition abrogates insulin receptor phosphorylation, but this effect is directly caused by changes in membrane biophysical properties, with the insulin receptor unable to translocate into detergent-resistant membranes (Park et al., 2012). Likewise, palmitic acid is not internalized by CerS2 null hepatocytes, caused by the inability of CD36 (also known as FAT) to translocate into detergent-resistant membranes before internalization (Park et al., 2014). The mechanism by which tumor necrosis factor α receptor-1 internalization is inhibited in CerS2 null hepatocytes has not been established (Ali et al., 2013), but we have suggested that this might be related to the altered recruitment of adaptor proteins required for clathrin-coated pit formation. Finally, altering the SL acyl chain composition also affects intracellular trafficking of connexin 32 (Park et al., 2013), again due to altered biophysical properties. Thus, SLs have direct effects on receptor internalization and trafficking but also indirect effects, such as those reported in the current study, whereby dysfunction of mitochondrial complex IV affects ROS levels and results in concomitant changes in pathways affected by ROS, such as endocytosis (Parry et al., 2008; Poumay and Ronveaux-Dupal, 1988). Whether similar changes in signaling pathways occur upon acute changes in SL levels (Morales et al., 2007) is currently unknown, but the current study highlights the need to pay attention to both direct effects of SLs on membrane biophysical properties and also to indirect effects caused by activation of downstream pathways, such as defective clathrin uncoating caused by oxidative stress.

Cortical astrocytes from 2- to 4-day-old WT and CerS2 null mice (Pewzner-Jung et al., 2010b) were isolated as described (Schildge et al., 2013). Astrocytes were grown to 95% confluence for 7-10 days, other cells were removed from the dishes by vigorous shaking (2 h, 200 rpm on an orbital shaker), and astrocytes were replated and grown to 95% confluence for up to 23 days. Cells were cultured in Dulbecco's modified eagle's medium (DMEM, Gibco™) containing 10% fetal calf serum (Gibco™) and 2% penicillin-streptomycin.

Astrocytes were cultured on 12- or 13-mm glass coverslips to a confluence of ∼70%. Cells were fixed by incubating with 3% paraformaldehyde in phosphate buffered saline (PBS) at room temperature for 15 min, rinsed three times with PBS for 4 min and blocked with 4% bovine serum albumin for 90 min. Cells were then either co-incubated for 1 h with anti-GFAP (for astrocytes, 1:1000; cat. no. 4560-0204, Bio-Rad) and anti-CD68 (for microglia, 1:1000; cat. no. MCA1957, Bio-Rad laboratories) antibodies, or with an anti-clathrin heavy chain (1:50; cat. no. CR796S, Cell Signaling) antibody. Cells were washed three times with PBS for 4 min and incubated with a secondary fluorescently labeled antibody (anti-rat, anti-mouse or anti rabbit; 1:250, Jackson Immunoresearch Labs) for 60 min. Cells were rinsed three times with PBS for 4 min, incubated for 10 min with Hoechst 33342 (1:5000, Molecular Probes) and then rinsed three times with PBS for 4 min. Glass coverslips were mounted on slides using Gel Mount (Molecular Probes, Eugene, Oregon, USA) and analyzed by fluorescence microscopy using an Olympus IX 81 FluoView 1000 microscope (Olympus, Tokyo, Japan).

Total RNA was isolated using the RNeasy mini kit (Qiagen) according to the manufacturer's instructions. cDNA synthesis was performed using a Verso cDNA kit (Thermo Scientific). Quantitative (q)PCR was performed as described previously (Vitner et al., 2016). The relative amounts of mRNA were calculated from the cyclic threshold values using hypoxanthine guanine phosphoribosyl transferase (HPRT) or TATA binding protein (TBP) for normalization. Primers are listed in Table S1.

Cells were homogenized in 20 mM HEPES-KOH, pH 7.2, 25 mM KCl, 250 mM sucrose and 2 mM MgCl₂ containing a protease inhibitor cocktail (Sigma). Protein content was determined using the BCA reagent (Pierce). Homogenates were incubated with 15 μM NBD-sphinganine (Tidhar et al., 2014) (Avanti Polar Lipids), 20 μM de-fatted bovine serum albumin (Sigma) and 50 μM C16:0-CoA (for CerS5 and CerS6), C18:0-CoA (for CerS1 and CerS4), C20:0-CoA (for CerS4) and C22:0-CoA, C24:0-CoA or C24:1-CoA (for CerS2) (Avanti Polar Lipids) for 5-40 min at 37°C. Lipids were extracted and separated by performing thin layer chromatography using chloroform:methanol:2 M NH₄OH (40:10:1, v/v/v) as the developing solvent. NBD-labeled lipids were visualized using a Typhoon 9410 variable mode imager and quantified using ImageQuantTL software (GE Healthcare).

SL analysis using liquid chromatography-coupled electrospray ionization tandem mass spectrometry (LC-ESI MS/MS) was performed using a PE-Sciex API 3000 triple quadrupole mass spectrometer and an ABI 4000 quadrupole-linear ion trap mass spectrometer (Shaner et al., 2009; Sullards et al., 2011).

WT and CerS2 null astrocytes were cultured in 96-well plates and grown overnight. Cells were incubated for various times with fluorescently labeled ligands [Dil–LDL (1:100), BODIPY–LacCer (1:100) from Life Technologies; Cy5–Tf (1:100) from Jackson Immunoresearch Labs]. Cells were washed three times with PBS to remove unbound ligand and incubated for 4 min with 500 mM NaCl in 200 mM acetic acid, pH 3.5, to remove Dil–LDL or to remove Cy5–Tf bound to the cell surface (Megias et al., 2000). For BODIPY–LacCer, ‘back exchange’ was performed by incubating cells with 5% de-fatted bovine serum albumin for 10 min. In some cases, cells were incubated with 100 μM H₂O₂ or 100 μM NAC for 4 h before addition of the fluorescent ligand. Internalized fluorescent ligand was quantified using a plate reader at the following wavelengths: Dil–LDL, excitation 528 nm, emission 590 nm; BODIPY-LacCer, excitation 485 nm, emission 528 nm; Cy5–transferrin, excitation 630 nm, emission 680 nm.

Western blotting (Zigdon et al., 2015) was performed with the following antibodies diluted in PBS and 0.1% Tween-20: rabbit pan anti-CHC (1:1000; cat. no. C4796S, Cell Signaling); rabbit pan anti-LDL receptor (1:1000, cat. no. AB30532), mouse anti-Hsc70 (1:1000, cat. no. ab2788), rabbit anti-Sp1 (1:1000, cat. no. ab13370), rabbit anti-3-nitrotyrosine (1:1000, cat. no. ab7048) antibodies, all obtained from Abcam (Cambridge, UK); rabbit anti-Tf receptor antibody (1:1000, cat. no. sc7087, Santa Cruz); and mouse anti-GAPDH antibody (1:5000, cat. no. MAB374, Millipore). Densitometry was performed using ImageQuant software (Amersham Biosciences).

Transcription factor binding site analysis was performed using the Genomatix Genome Analyzer (GGA) MatInspector program (Cartharius et al., 2005), Matrix Library Version 9.0, searching with the V$SP1F matrix.

ChIP was performed as described previously (Tarcic et al., 2016). Cells were fixed in 0.1 volume of formaldehyde (11% stock solution) followed by 0.06 volumes of 2.5 M glycine. Immunoprecipitation was performed using 3 µg of antibody in 0.1% Brij-35 solution. DNA was purified using a Qiagen PCR purification kit (Qiagen) followed by qPCR.

ROS were measured using chloro-methyl 6-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (cmH₂DCF-DA) (Invitrogen). WT and CerS2 null astrocytes were incubated with 100 μM cmH₂DCF-DA or PBS for 30 min at 37°C and then washed three times with PBS. Conversion of cmH₂DCF-DA to cmDCF was determined using an Eclipse iCyt flow cytometer equipped with 488 nm solid state air cooled lasers, with 25 mW on the flow cell and with standard filter set-up; cmDCF was measured in the green channel with an excitation of 488 nm and emission of 525±50 nm.

Activities of mitochondrial respiratory chain complex IV and citrate synthase, a mitochondrial marker, were determined in cell homogenates as previously described (Saada et al., 2004). Measurements were performed using a double-beam spectrophotometer (UVIKON 930, Secomam, France).

Confocal microscopy was performed using an Olympus IX 81 FluoView 1000 microscope and a UPLSAPO 60× objective. Images were processed and analyzed using FV-1000 software (Olympus, Tokyo, Japan).

We thank Mrs Shiri Graff from the laboratory of Dr Assaf Vardi for help with the FACS analysis of ROS and Mr Vladimir Kiss for assistance with high-resolution fluorescence and confocal microscopy.