Identification and characterization of protein phosphatase 2C activation by ceramide.

Ceramide is a bioactive sphingolipid with many associated biological outcomes, yet there is a significant gap in our current understanding of how ceramide mediates these processes. Previously, ceramide has been shown to activate protein phosphatase (PP) 1 and 2A. While continuing this line of work, a late fraction from a Mono-Q column was consistently observed to be activated by ceramide, yet PP1 and PP2A were undetectable in this fraction. Proteomic analysis of this fraction revealed the identity of the phosphatase to be PP2Cγ/PPM1G. This was consistent with our findings that PP2Cγ 1-eluted in a high salt fraction due to its strongly acidic domain, and 2-was insensitive to okadaic acid. Further characterization was performed with PP2Cα, which showed robust activation by C6-ceramide. Activation was specific for the erythro conformation of ceramide and the presence of the acyl chain and hydroxyl group at the first carbon. In order to demonstrate more physiological activation of PP2Cα by ceramide, phospho-p38δ was utilized as substrate. Indeed, PP2Cα induced the dephosphorylation of p38δ only in the presence of C16-ceramide. Taken together, these results show that the PP2C family of phosphatases is activated by ceramide, which may have important consequences in mediating the biological effects of ceramide.

with 200 U PKA, 1 mM ATP, 50 Ci of either 32 P ␥ -or 33 P ␥ -ATP, 10 mM MgCl 2 , 5 mM DTT, 10 mM ␤ ME in a fi nal volume of 500 l of 50 mM Tris-HCl buffer pH 7.4 for 2 h at 37°C. This was then precipitated with 170 l TCA ( Ϫ 20°C) on ice for 10-30 min and centrifuged at 15,000 g for 15 min. The supernatant was discarded and the pellet was resuspended in 500 l of 40 mM Tris-HCl pH 7.4. This was repeated a minimum of three times to effectively remove unincorporated ATP.

Phosphatase activity assays
Phosphatase assays were performed as described previously ( 11 ) with minor modifi cations. Briefl y, ceramides and other lipids were dissolved in either ethanol or dodecane-ethanol (2%:98% v/v), which were added to assay buffer (50 mM Tris pH 7.4, 20% glycerol, 1 mM MnCl 2 , 100 mM NaCl, 1 mM DTT) not exceeding 1% of total volume and preincubated for 5 min at 30°C. For testing activity from FPLC fractions, 2-5 l of each fraction was incubated for 15 min (phenyl sepharose fractions) or 1 h (Mono-Q fractions) with radiolabeled MBP. For assays with PP2C ␣ , 5-10 ng of enzyme diluted in assay buffer was incubated in assay buffer for 30 min. Termination of the reaction was achieved by adding 375 l of cold 55% TCA followed by 275 l of 0.9 mg/ml BSA. Tubes were capped, vortexed, and incubated for 15 min on ice. Samples were then centrifuged for 30 min at 4°C from which 250 l of the supernatant was transferred to scintillation vials for measuring radioactivity as a readout of hydrolyzed phosphate. Statistical and kinetic analysis was performed with Prism/GraphPad using nonlinear regression analysis for calculating kinetic paramters.

Protein staining and Western blotting
Fractions were prepared for SDS-PAGE by freezing 0.3 ml at Ϫ 80°C and subsequent lyophilization overnight. Samples were then resuspended in Laemmli buffer and run on SDS-PAGE for total protein staining or Western blots on nitrocellulose membranes and were incubated overnight with primary antibodies in 5% nonfat milk in PBS/ 0.1% Tween 20. Membranes were washed in PBS/ 0.1% Tween 20 at least three times, incubated with secondary antibodies, washed again, and then developed in ECL chemiluminescence reagent.

Proteomic analysis
Fractions of interest were precipitated by adding 0.9 ml cold 100% acetone to 0.3 ml of each fraction and incubated at Ϫ 20°C overnight. Samples were centrifuged for 20 min at 45,000 g at 4°C. Supernatants were aspirated and then pellets were dried and stored. Acid-dissolved Trypsin Gold was diluted to 20 g/ml and added to the solubilized pellet and incubated overnight at 37°C for full digestion. The next day, solution was dried in SpeedVac for 2-4 h in preparation to full dryness then resolubilized with 50% acetonitrile/5% at room temperature for MS analysis.
An instrument set of LC/ESI/MS/MS on a linear ion trap mass spectrometer (LTQ, Thermo Finnigan) coupled to an LC Packings nano-LC system was used to analyze the trypsin-digested sample. For reversed phase LC column, a 15-cm × 75 m C-18 (packed by the Proteomics Core Facility, Medical University of South Carolina) was utilized with a 60 min gradient from 2% acetonitrile, 0.1% formic acid to 60% acetonitrile, 0.1% formic acid at a fl ow rate of 200 nl/min. A between samples blank was analyzed to limit carryover. In the mass spectrometer, an MS/MS was performed on all ions above an ion count of 1000 using datadependent analysis. Dynamic exclusion mode was set to exclude ions from MS/MS selection for 3 min after being selected two times in a 30 s window. identifi ed PP2C ␥ /PPM1G in a high-salt-eluting fraction from a Mono-Q column as ceramide activated. Further characterization was performed on PP2C ␣ , the canonical PP2C isoform, showing activation in a stereospecifi c and selective manner. This work has potential impact on the current understanding of the molecular mechanisms by which ceramide mediates biological outcomes, implicating PP2C ␣ as a potential target of ceramide signaling.

Fractionation of endogenous phosphatases
Fractionation was performed as described previously ( 24 ). In order to identify novel targets of ceramide-activated phosphatases from human sources, A549 lung cancer cells were used, and lysates were used immediately after lysis, compared with previous work which used frozen tissue. Likewise, frozen rat brains were used as a source of starting material to be able to compare with previous work and corroborate fi ndings from A549 cells. Total homogenate from A549 cells ( ‫ف‬ 10 T150 fl asks) or one rat brain was prepared by nitrogen cavitation or passing through a tightfi tting 7 ml Wheaton dounce in lysis buffer containing 50 mM Tris HCl pH 7.4, 1 mM EDTA, 1 mM EGTA, 1 mini-protease inhibitor tablet/10 ml (Roche, EDTA free). Homogenate was cleared by centrifuging at 1,000 g for 5 min to remove debris and nuclei. Supernatant was transferred to high-speed tubes and centrifuged at 100,000 g for 1 h. All steps were performed on ice or at 4°C.
Soluble lysate was mixed with 5 M NaCl suffi cient to bring to a fi nal concentration of 1.5 M NaCl. This was loaded onto a phenyl sepharose column and eluted by decreasing salt concentration to 0 M NaCl. Fractions were tested for phosphatase activity in the absence or presence of C 6 -ceramide. Aliquots were also stored for total protein staining or immunoblotting after SDS-PAGE. Fractions eluting from ‫ف‬ 0.5 M through 0 M NaCl were pooled (0.5 ml/fraction), diluted to 50 mM NaCl (50 ml maximum), and then loaded onto a Mono-Q column. This was then eluted by a linear gradient of 50 mM NaCl to 500 mM NaCl. Fractions were then tested for phosphatase activity in the absence or presence of C 6 -ceramide and then aliquotted for Coomassie staining and immunoblotting.

Radiolabeling of MBP
MBP was phosphorylated by PKA as described previously ( 11,24 ). Briefl y, in two identical reactions, 1 mg MBP was incubated with the presence of PP2A according to immunoblots for the catalytic subunit ( Figs. 1 and 2 ). As known previously, this activity was stimulated by ceramide roughly two-fold, with particular fractions demonstrating unique sensitivity suggesting specifi c ceramide sensitive targets, ruling out that this was a general effect. Fractions corresponding to 0.5 M NaCl and later were pooled, diluted, and loaded on a Mono-Q column.
Phosphatase activity derived from A549 eluting from the Mono-Q displayed three peaks of activity ( Fig. 1A, right panel. Peak I-fr 72, Peak II-fr 77-78, Peak III-fr 86). The fi rst two peaks correlated with the presence of PP2A by immunoblot, with peak II being clearly activated by ceramide ( Fig. 2A, right panel). The two different pools of PP2A are likely a result of specifi c heterotrimers with unique B subunits ( 24 ) conferring differing anionic character. Interestingly, a third peak of low basal activity was observed that lacked PP2A yet was consistently activated by C 6 -ceramide. None of the PP1 isoforms ( ␣ / ␤ / ␥ ) was detectable by immunoblot (data not shown), suggesting a previously undescribed ceramide-activated phosphatase.
In order to determine whether this phosphatase was in the PPP family, fractions representing peak II and III off the Mono-Q from A549 were analyzed for their sensitivity to OA and ceramide. OA strongly inhibited the PP2A-rich peak II as expected and largely blocked the ceramide activation. In contrast, peak III, which displayed low basal activity, was very sensitive to ceramide stimulation but insensitive The MS/MS data search was performed against mammalian databases using Thermo Finnigan Bioworks 3.3 software. Data search was also set to account for posttranslational modifi cations of methionine oxidation, carboxamidomethylation of cysteines, and phosphorylation of serine, threonine, and tyrosine.

Dynamic light scattering
Varying concentrations of C 16 -ceramide were prepared in dodecane/ethanol (2%:98%). This solution was added to activity assay buffer 1:100 (v/v) and analyzed by dynamic light scattering for measurement of particle size using a PDDLS/ Cool Batch 40T/ PD2000 DLS. Settings for measurement included 20°C, 0.01769 P and a refraction of 1.35749.

Identifi cation of PP2C ␥ as ceramide-activated
In an attempt to characterize the in vitro activation of Ser/Thr phosphatases by ceramide, the approach developed by Galadari et al. ( 24 ) was employed. Soluble lysate from A549 cells or rat brain was fractionated fi rst on a phenyl sepharose hydrophobic exchange column and subsequently on a Mono-Q anionic exchange column. After each column, fractions were tested for phosphatase activity in the presence or absence of C 6 -ceramide (ethanol vehicle) using radiolabeled 32 Pi-MBP as a substrate. Basal phosphatase activity eluting from the phenyl sepharose column is characterized by a large peak, which generally correlated Fig. 1. Ceramide-activation of serine/threonine phosphatase activity in fractions from phenyl sepharose and Mono-Q columns. Soluble lysate from A549 (A) or rat brain (B) was fractionated over phenyl sepharose columns with a gradient from 1.5 M to 0 M NaCl. Fractions were subsequently assayed for phosphatase activity as described in Materials and Methods in the absence or presence of 15 M C 6 -ceramide. Fractions consisting of 0.5 M to 0 M NaCl were pooled and diluted to ‫ف‬ 50 mM NaCl, and then loaded onto a Mono-Q column. Proteins were eluted by a linear gradient from 50 mM NaCl to 500 mM NaCl. Subsequently fractions were assayed in the absence of presence of 15 M C 6 -ceramide. Results are representative of at least three independent experiments. family, due to diffi culties in expressing active full-length PP2C ␥ . To fi rst test whether PP2C ␣ was also activated by ceramide in vitro, purifi ed recombinant PP2C ␣ was assayed in the presence of 30 M C 6 -ceramide delivered in ethanol or dodecane/ethanol (2%/98%) as a delivery system because this is the vehicle of choice for long chain ceramides in vitro ( Fig. 3A , B ). Both systems resulted in activation of PP2C ␣ . Similar results were obtained with other long chain and very long chain ceramides (data not shown). Taken together, the data suggest ceramide was able to strongly activate PP2C ␣ whether delivered with ethanol or dodecane/ethanol.
A dose response with C 16 -ceramide was performed on PP2C ␣ using dodecane/ethanol revealing an EC 50 of activation of 5.3 M ( Fig. 3C ). The lack of linearity with increasing ceramide concentrations on PP2C ␣ activity suggested a dynamic process was occurring between 4 and 10 M. To investigate whether this activation was simply a result of particle formation in this range, dynamic light scattering was performed under similar conditions on dodecane/ethanol with increasing C 16 -ceramide concentrations (supplementary Fig. I). Dodecane/ethanol alone at 1% v/v in activity assay buffer formed particles and upon addition of ceramide, an increase in particle size was observed, which modestly continued for 5 and 10 M. Due to a lack of a dynamic change in particle size in the range where activation of PP2C ␣ was seen, it is unlikely that activation by ceramide is simply the result of lipid particle formation.

Effect of ceramide on kinetic parameters of PP2C ␣
Next, the effects of ceramide on kinetic parameters of PP2C ␣ were evaluated. C 16 -ceramide strongly increased V max while having a minor but notable increase in K m ( Fig. 4A , Table 1 ). Overall, there was an increase in k cat /K m to OA in the absence or presence of ceramide ( Fig. 2C ). These results further argued against the possibility that the identity of the phosphatase in peak III was a previously characterized ceramide-activated protein phosphatase.
Separately, corresponding fractions to peaks I, II, and III were observed from the Mono-Q in rat brain although separation between peaks was less than with A549 lysate ( Fig. 1B, right panel). In order to ascertain the identity of the phosphatase in peak III, LC/MS/MS proteomic analysis was performed on fraction 83 from the Mono-Q using rat brain ( Fig. 1A, right panel). Two peptide fragments were identifi ed with high confi dence matching to PP2C ␥ / PPM1G (XC score of 20.27, delta CN of 0.51 and 0.33 for each peptide), which were specifi c to this fraction, raising the possibility that PP2C ␥ was responsible for the ceramide activation of phosphatase activity in peak III. Total protein staining of SDS-PAGE was unable to detect a band in the expected molecular weight range for PP2C ␥ in this fraction most likely due to low abundance (data not shown). The PP2C/PPM family of protein phosphatases is metal ion-dependent and insensitive to OA, fi tting with the fi nding that peak III was insensitive to OA. Moreover, PP2C ␥ possesses a strongly acidic domain of which it is the only Ser/Thr phosphatase with this domain. This corroborates the fi nding that peak III corresponds to a high-salt-eluting fraction due to its strong acidic domain yielding a predicted overall pI of 4.36. This is consistent with previous work when it was originally cloned and purifi ed from a Mono-Q column where it also eluted at a similar NaCl concentration ( 19 ).

Activation of PP2C ␣ by short chain and natural ceramides
Further characterization of PP2C activation by ceramide was developed using PP2C ␣ , the canonical isoform of this for its substrate or cofactors but directly activates the enzyme, likely by allosteric change in the conformation of PP2C ␣ .

Lipid specifi city of activation
Ceramide has chiral centers at carbons 2 and 3, yielding four stereoisomers. Evaluation of the enantiomer and by 3.1-fold. In order to determine whether ceramide regulates the affi nity for the metal cofactor, PP2C ␣ activity was measured at increasing concentrations of MnCl 2 in the absence or presence of 10 M C 16 -ceramide ( Fig. 4B ). C 16 -ceramide resulted in an expected increase in V max with no signifi cant change in K A for MnCl 2 ( Table 2 ). This suggests that ceramide may not regulate the affi nity of PP2C ␣  dephosphorylation of p38 ␦ ( Fig. 6A , B ). An additional dose response was performed with C 16 -ceramide from 100 nM to 1 M showing effects in the high nanomolar range ( Fig. 6C ). Additionally, a similar lipid specifi city was seen with p38 ␦ as compared with MBP where there was a preference for the erythro conformation and sphingoid backbone ( Fig. 6D ). These results demonstrate that ceramide allows PP2C ␣ to dephosphorylate p38 ␦ .

DISCUSSION
In this study, PP2C was discovered to be a novel ceramide-activated PP. Using A549 human lung adenocarcinoma cells and rat brain as sources of soluble lysate, a fraction, with relatively low basal activity and devoid of known ceramide-activated protein phosphatases , was activated by C 6 -ceramide and later identifi ed by LC/MS/MS from rat brain to be PP2C ␥ . In order to further study this interaction, the canonical PP2C isoform, PP2C ␣ /PPM1A, was used. Both PP2C ␥ and PP2C ␣ were activated under similar conditions by C 6 -ceramide delivered in ethanol. C 16 -ceramide resulted in a robust increase in catalytic activity of PP2C ␣ , where this activation was stereospecifi c. Lastly, PP2C ␣ induced the dephosphorylation of p38 ␦ , a known physiologic substrate of PP2C, only in the presence of C 16 -ceramide.
This work demonstrates that PP2C ␣ is robustly and specifi cally activated by ceramide in vitro. PP2C ␣ was strongly activated by C 6 -ceramide whether delivered in ethanol alone or dodecane/ethanol, demonstrating that dodecane itself was not necessary for activation but possibly enhancing delivery. Moreover, dodecane/ethanol is likely to be more physiological where lipid surfaces or particles are being made; however, the exact nature of this structure is unknown.
Within the sphingoid structure, there was remarkable specifi city for ceramide on the activation of PP2C ␣ . Structure/function studies revealed a requirement for the presence of the acyl chain, primary hydroxyl group, and erythro conformation. Moreover, the lack of activation by saturated or unsaturated fatty acids and the diminished activation by DAG compared with ceramide demonstrate the ceramide/PP2C ␣ interaction to be rather specifi c. This raises the possibility that there is a unique cavity or pocket in PP2C ␣ able to sense and respond to ceramide, possessing a hydrophobic pocket with polar residues in the core able to interact with the hydroxyl group and amide of ceramide.
Due to the lack of correlation of activity to particle size, ceramide more likely binds specifi cally to PP2C ␣ , provoking an allosteric change rather than nonspecifi c particle formation resulting in activation. This is further supported by the evidence that threo stereoisomers did not activate PP2C ␣ ( Fig. 5A ). Standard lipid/protein overlay assays were unsuccessful in detecting binding between ceramide and PP2C ␣ , an assay that is known not to be well suited for neutral lipids (data not shown). Additionally, conventional methods of delivering lipids, liposomes/micelles, did not result in strong changes in activity (data not shown), suggesting diastereomers of D-erythro -C 16 -ceramide demonstrated that the erythro conformation (both D-erythro -and L-erythro ) was required for activation of PP2C ␣ , and the threo conformation inhibited activity ( Fig. 5A ).
Other structural variants of ceramide were evaluated to assess the structural requirements for activation of PP2C ␣ by ceramide. Sphingosine, which lacks the acyl chain, resulted in inhibition of PP2C ␣ ( Fig. 5B ). The primary hydroxyl at position 1 proved to be essential for activation due to 1-deoxy-C 16 -ceramide having no effect on activity ( Fig. 5B ). Ceramide possesses a unique trans double bond at position C4-5, which appeared to not be essential for activation, as dihydro-C 16 -ceramide was still capable of activation ( Fig. 5B ). This is distinct from the near lack of activation of PP2A and PP1 by dihydro-ceramides ( 11 ).
It has been reported that unsaturated fatty acids are able to activate PP2C ␣ at high micromolar-to-millimolar concentrations in DMSO ( 25 ). In our system, the saturated and unsaturated fatty acids, palmitate and oleate, respectively, both showed inhibition of activity at 10 M ( Fig. 5C ) as well as at higher concentrations (data not shown). In order to assess whether this activation is specifi c to the sphingoid backbone, the glycerolipid, dipalmitoylglycerol (DAG) was compared with C 16 -ceramide. DAG also displayed activation but was signifi cantly less than C 16 -ceramide ( Fig. 5C ). Overall, the results demonstrate stereochemical and structural specifi city and selectivity in regulation of PP2C by ceramides demonstrating the headgroup of ceramide to be the critical region for activation.

Ceramide-dependent dephosphorylation of p38 ␦ by PP2C ␣
The stress-activated p38 MAP kinase is a well-known in vivo and in vitro substrate for PP2C ␣ ( 26,27 ). Lack of phosphorylation at Thr180 in the activation loop of p38 was suffi cient to completely inactivate p38 ␣ ( 28,29 ), implicating this as a possible mechanism for regulation. Previous work from our laboratory has shown that p38 activation is regulated by endogenous ceramide where the p38 ␦ isoform is preferentially targeted ( 16,17 ). Therefore, using phosphorylated p38 ␦ as a phospho-substrate, ceramide activation of PP2C ␣ was evaluated at varying doses in dodecane/ethanol. PP2C ␣ alone did not dephosphorylate p38 ␦ noticeably; however, the presence of low micromolar concentrations of ceramide caused a robust is not understood, and further studies are needed in this regard. Finally, PP2C ␣ was virtually unable to dephosphorylate p38 ␦ in the absence of ceramide (even at high concentrations a unique lipid delivery requirement for ceramide to be bioactive in this context such as the requirement for more soluble monomeric ceramide. Taken together, the data indicate the exact nature of the ceramide/PP2C ␣ interaction   6. Ceramide-dependent dephosphorylation of phospho-p38 ␦ . A: PP2C ␣ activity was measured as described in Materials and Methods using recombinant phospho-p38 ␦ as substrate with increasing concentrations of C 16 -ceramide in dodecane/ethanol (2%/98%). Increasing concentrations of C 16 -ceramide were incubated with recombinant p38 ␦ with or without PP2C ␣ . B: Phosphorylated p38 ␦ was quantifi ed from the upper panel including PP2C ␣ using ImageJ software. Data represent mean ± SE of four separate determinations. One-way ANOVA statistical analysis with Bonferroni's post test, * P < 0.01, ** P < 0.001. C: Concentrations of C 16 -ceramide from 100nM to 1 M were tested on PP2C ␣ using phospho-p38 ␦ as substrate as described in Materials and Methods. Levels of phospho-p38 and p38 ␦ were quantifi ed and expressed as fold compared with control. D: PP2C ␣ was assayed using phospho-p38 ␦ as substrate in the presence of 10 M D-erythro , D-threo , L-erythro , or L-threo -C 16 -ceramide or dipalmitoyl-glycerol. The ratio of phospho-p38 to p38 ␦ is expressed as fold change compared with control. of PP2C ␣ , which are known to have high activity toward MBP). However, in the presence of low micromolar concentrations of ceramide, there was almost complete dephosphorylation of p38 ␦ . This suggests that with a more physiological phospho-substrate, activation by ceramide may be a direct and robust mechanism for regulation.
Accumulating evidence is emerging that implicates PP2C ␣ in responses that are known to be regulated by sphingolipids, suggesting cellular connections. First, several lines of work demonstrate that p38 is regulated by the salvage pathway of sphingolipids where ceramide is likely a negative regulator of p38 phosphorylation ( 16 ). Inhibition of ceramide-producing enzymes, acid ␤ -glucosidase and ceramide synthase, resulted in enhanced phosphorylation of p38, and conversely, exogenous treatment of cells with C 6ceramide caused a decrease in p38 phosphorylation induced by phorbol 12-myristate 13-acetate in MCF-7 cells ( 16,17 ). PP2C ␣ is a key phosphatase of this pathway, known to dephosphorylate Thr180 of p38. This model is consistent with our in vitro data suggesting that ceramide activation of PP2C ␣ may be an acute mechanism of negative regulation of p38 signaling. Second, SMAD 2/3 is regulated by sphingosine-1-phosphate and dihydro-sphingosine-1-phosphate ( 30 ). Interestingly, these two lipids have opposite effects on SMAD phosphorylation, which may be mediated through PP2C ␣ , an essential SMAD2/3 phosphatase. The role of ceramide in this context has not been evaluated where it is possible that ceramide is a negative regulator of SMAD phosphorylation via PP2C ␣ activation.
The fi nding that ceramide activates PP2C in addition to PP1/PP2A raises some considerations. As to the target of ceramide, this suggests that the catalytic subunit of the phosphatase is the target based on the monomeric nature of PP2C compared with the more complicated scenario of PP1/PP2A with multiple scaffold and regulatory subunits. Interestingly, despite the lack of sequence of homology between the family of phosphatases, there appears to be a similar structural fold between them, which implies that ceramide may be interacting within this region of the enzyme. However, it is entirely plausible that ceramide has different targets in the context of PP2C versus PP1/PP2A.
In conclusion, this study demonstrates for the fi rst time a connection between ceramide and the OA-insensitive phosphatase family, PP2C. Further study is necessary to probe the physiological signifi cance of this fi nding in the context of stress signaling.