Address correspondence and reprint requests to Anthony D. Postle, Division of Infection, Inflammation and Immunity, School of Medicine, Southampton General Hospital, Southampton SO16 6YD, UK; E-mail: A.D.Postle@soton.ac.uk or to Cesare Montecucco, Department of Biomedical Sciences, University of Padova, Via G. Colombo 3, 35121 Padova, Italy; E-mail: firstname.lastname@example.org
Snake pre-synaptic phospholipase A2 neurotoxins paralyse the neuromuscular junction by releasing phospholipid hydrolysis products that alter curvature and permeability of the pre-synaptic membrane. Here, we report results deriving from the first chemical analysis of the action of these neurotoxic phospholipases in neurons, made possible by the use of high sensitivity mass spectrometry. The time–course of the phospholipase A2 activity (PLA2) hydrolysis of notexin, β-bungarotoxin, taipoxin and textilotoxin acting in cultured neurons was determined. At variance from their enzymatic activities in vitro, these neurotoxins display comparable kinetics of lysophospholipid release in neurons, reconciling the large discrepancy between their in vivo toxicities and their in vitro enzymatic activities. The ratios of the lyso derivatives of phosphatidyl choline, ethanolamine and serine obtained here together with the known distribution of these phospholipids among cell membranes, suggest that most PLA2 hydrolysis takes place on the cell surface. Although these toxins were recently shown to enter neurons, their intracellular hydrolytic action and the activation of intracellular PLA2s appear to contribute little, if any, to the phospholipid hydrolysis measured here.
The venom of many Australian and Asiatic Elapid snakes is highly poisonous and the majority of the symptoms of envenomation in humans is the result of the action of neurotoxins endowed with phospholipase A2 activity (PLA2) activity, which are prominent components of these venoms (Chen and Lee 1970; Connolly et al. 1995; Harris et al. 2000; Prasarnpun et al. 2005). Human envenomation is a major health concern in many countries of the world (Gutiérrez et al. 2006). These neurotoxins are abbreviated here as snake pre-synaptic PLA2 neurotoxins (SPANs). The role of their PLA2 enzymatic activity in the blockade of the neuromuscular junction (NMJ) has been long debated (Rosenberg 1997; Kini 2003; Gutiérrez et al. 2006), but recent results demonstrated that their phospholipid hydrolysis products, lysophospholipids (LysoPLs) and fatty acids (FAs), are sufficient to cause NMJ paralysis with the associated pathological changes (Rigoni et al. 2005; Caccin et al. 2006). Therefore, to define the pathological action of SPANs it is necessary to analyse their PLA2 activity in vivo as a function of time. Recent advances in mass spectrometry (MS) of lipids (Han and Gross 2003; Pulfer and Murphy 2003; Postle et al. 2007; Davis et al. 2008; Wilensky et al. 2008) have enabled us to analyse PLA2 hydrolysis products with high sensitivity and specificity. Electrospray ionization (ESI)-MS, coupled with diagnostic tandem MS/MS scans, permits routine comprehensive characterization of membrane lipids from as few as 106 cells.
It is still impossible to analyse quantitatively the lipid products released by SPANs at the NMJ, their principal target in humans, for obvious technical reasons. However, it is well established that SPANs are highly toxic when injected into the CNS (Gandolfo et al. 1996; Kolko et al. 1999) and act on isolated brain-derived preparations (Rehm and Betz 1982; Nicholls et al. 1985; Rugolo et al. 1986). Therefore, data obtained with cultured CNS neurons were relevant and were as close to the in vivo situation as is currently experimentally possible. Methods are available to maintain many CNS neurons in primary cultures, but these cultures are mixtures of neurons and glial cells except for the granular neurons of the cerebellum (CGNs, cerebellar granule neurons), wherein cultures consist almost entirely of neurons (Lasher and Zaigon 1972; Levi et al. 1984). We have shown previously that SPANs are very active on these neurons (Rigoni et al. 2004). Using a pure neuronal culture is essential to achieve consistent and reliable MS analysis of changes in lipid compositions, as the presence of a large component of SPAN-resistant glial cells would dilute the changes induced by the toxins. Moreover, glial cells and neurons will have distinct and different compositions of membrane lipid, which will further complicate the MS analysis. CGN neurons are highly sensitive to SPANs and develop a well-defined bulging at axon and dendrite terminals within few minutes from toxin addition; such morphological alteration is accompanied by cytosolic calcium increase at nerve terminals and glutamate release from neurons (Rigoni et al. 2004, 2007). These effects are mimicked by the addition of an equimolar mixture of the PLA2 hydrolysis products, lysophosphatidylcholine (LysoPC) + oleic acid, indicating that these molecules are the biochemical mediators of SPAN action (Rigoni et al. 2005; Caccin et al. 2006). Of the two lipid molecules, LysoPC was shown to be most effective (Caccin et al. 2006, 2009). We have shown previously that LysoPC is by far the major class of LysoPL released by two SPANs (Rigoni et al. 2005), but a comparative and detailed analysis of their kinetics of action including the other two major plasma membrane phospholipids [phosphatidylethanolamine (PE) and phosphatidylserine (PS)] is still lacking. To obtain results of rather general interest here, we have challenged CGN primary cultures with four different neurotoxic SPANs that differ in terms of quaternary structure and in vitro PLA2 activity: notexin (Ntx), β-bungarotoxin (β-Btx), taipoxin (Tpx) and textilotoxin (Tetx). Ntx is a 14 kDa monomer, β-Btx is a heterodimer with one 14 kDa active PLA2 subunit, Tpx is a trimer of similar 14 kDa subunits one of which is PLA2 active, and Tetx consists of five or six similar 14 kDa subunits, only one of which is enzymatically active (Montecucco and Rossetto 2008; Aquilina 2009). The results obtained provide the kinetics of phospholipid hydrolysis in a homogeneous population of neurons and strongly indicate the outer layer of the plasma membrane as their major site of phospholipid hydrolysis.
Materials and methods
Notexin, Tpx and Tetx were purchased from Venom Supplies (Tanuda, South Australia); β-Btx was obtained from Sigma (St. Louis, MO, USA). Their purity was controlled by sodium dodecyl sulfate–polyacrylamide gel electrophoresis. Aristolochic acid was supplied by BIOMOL Research Laboratories (Plymouth Meeting, PA, USA) and ionomycin by Calbiochem (San Diego, CA, USA). Phospholipid internal standards were obtained from Avanti Polar Lipids (Alabaster, AL, USA).
Rat CGNs were prepared from 6-day-old Wistar rats as previously described (Levi et al. 1984) and plated at 2 × 106 cells per 35-mm Petri dish. Cells were used 6 days after plating. NSC34 cells were maintained in Dulbecco’s modified Eagle’s medium with sodium pyruvate, supplemented with 10% fetal bovine serum. Cells were plated at 4 × 104 cells per 35-mm Petri dish and differentiated for 5 days with 5% fetal bovine serum and 10 μM retinoic acid before intoxication.
The enzymatic activity of the four SPANs was measured using the 1,2-dithio analogue of di-heptanoyl PC as substrate (Cayman Chemicals, Ann Arbor, MI, USA). The hydrolysis of the thio ester bond at the sn-2 position by PLA2 generated free thiols that interacted with 5,5′-dithiobis-(2-nitrobenzoic acid), leading to an increase in absorbance at 405 nm (Reynolds et al. 1992). ΔA405 was measured with a Packard SpectraCount spectrophotometer (Packard, Chicago, IL, USA).
Cerebellar granule neurons (6 days in vitro) grown on 35-mm polylysine-coated Petri dishes (2 × 106 cells/dish) were extensively washed with pre-warmed Krebs–Ringer buffer (125 mM NaCl, 5 mM KCl, 1.2 mM MgSO4, 2 mM CaCl2, 1.2 mM KH2PO4, 6 mM glucose and 25 mM HEPES-Na, pH 7.4) and then incubated for 0, 5, 10, 20, 40 and 60 min with each toxin (Ntx, β-Btx, Tpx and Tetx) at 6 nM concentration in 1 mL Krebs-Ringer buffer. When needed, a 20 min pre-incubation with 1 μM ionomycin was performed. In some cases, cells were pre-treated with 50 μM aristolochic acid (AA) for 30 min before 60-min incubation with the toxins. All treatments were performed at 37°C. After intoxication, cells were broken in cold methanol, collected and pellets were frozen and kept at −20°C until analysis. Total lipids were extracted from cell pellets using CHCl3–CH3OH (2 : 1) after addition of the internal standards dimyristoylPC (2 nmol), dimyristoylphosphatidylethanolamine (0.8 nmol), dimyristoylPS (0.4 nmol), dimyristoylphosphatidic acid (0.2 nmol), dimyristoylphosphoglycerol (0.4 nmol) and heptadecanoyl LysoPC (0.4 nmol). The chloroform fraction was dried under a gentle nitrogen stream and stored at −80°C until analysis. ESI-MS was performed on a Micromass Quatro Ultima triple quadrupole mass spectrometer (Micromass, Wythenshaw, UK) equipped with an ESI interface. Dried lipid extracts were dissolved in CH3OH : CHCl3 : H2O : NH3 (7 : 2 : 0.8 : 0.2, v : v) and injected into the mass spectrometer at a flow rate of 5 μL/min. Phosphatidylcholine (PC) and PE species were preferentially detected using positive ionization while PS was quantified under negative ionization conditions. Following fragmentation with argon gas, PC and LysoPC molecules were quantified from precursor scans of m/z 184, PE and lysophosphatidylethanolamine (LysoPE) molecules by neutral loss scans of m/z 141 and PS and Lyso PS molecules by neutral loss scans of m/z 87. Data were acquired and processed using MassLynx NT software (Micromass, Wynthenshow, UK). After conversion to centroid format according to area, correction for 13C isotope effects and for reduced response where appropriate of tandem MS/MS scans with increasing m/z values, the phospholipid species were expressed as percentages of their respective totals present in the sample.
CGNs phosphatidylcholine hydrolysis by SPANs
Figure 1 shows the time–courses of LysoPC production by the four SPANs included in this study; toxin concentrations and time periods of analysis were chosen based on both the morphological and functional observations described above. Ntx displayed the fastest PLA2 activity, closely followed by β-Btx; Tetx appeared to be a less potent PLA2 whereas Tpx displayed an intermediate activity. While the PLA2 activity of Ntx and β-Btx reached a plateau after 30–40 min from addition, the two multisubunit SPANs, Tpx and Tetx, were still active after 60 min and even after a longer incubation (2 h, results not shown). The present analysis was limited to 1 h because all these neurotoxins had paralysed the NMJ and bulged neurons in culture within this time limit (see Table 1). To ascertain whether the plateau reached by Ntx was because of its inactivation or inhibition, an equal amount of fresh toxin was added after 60 min from the first addition (see inset in Fig. 1), but no further production of LysoPC was observed. The PLA2 activity of Ntx appears therefore to be inhibited, although we do not know if this is because of inhibition by the hydrolysis products, or lower accessibility of the phospholipid substrates or other reasons.
Table 1. Comparison between the time course of LysoPC production and paralysis of the mouse NMJ by the four different SPANs
Times at which 50% and 90% of lysoPC production (with respect to the amount obtained after 1 h of neurons exposure to the indicated neurotoxin) and of NMJ transmission blockade were determined and are reported in minutes from intoxication.
aToxin concentration in mouse NMJ paralysis experiments was 1 μg/mL for Ntx, Tpx and Tetx, 3 μg/mL for β-Btx.
The amount of lipids released after 1 h of intoxication was similar in the four cases with about 17% of total cellular PC hydrolysed by Ntx, 19% by Tpx and 13% by β-Btx and Tetx. Clearly, this was a sizeable proportion of the total cell lipids. MS characterization of phospholipids composition of cerebellar neurons gave the following result: PC (59% of total phospholipids), PE (25%) and PS (15%). The time–course of LysoPC production was comparable with that of the paralysis induced by the same neurotoxins in the mouse hemidiaphragm, as shown by the comparison of their t50% and t90% reported in Table 1. In all cases but one, the time–course of phospholipid hydrolysis was lower than the corresponding one for NMJ paralysis and this might be taken as an additional evidence in favour of a consequentiality between PLA2 activity and blockade of the nerve terminal by these snake neurotoxins.
SPAN hydrolysis of phosphatidylethanolamine and phosphatidylserine in CGN neurons
Figure 2a shows that LysoPE was also produced by the four SPANs with time–courses comparable with that of LysoPC, but to lower extents. There was a difference among the four SPANs, with Ntx showing a much higher activity toward PE than the other three neurotoxins. In any case, the ratio LysoPC/LysoPE calculated after 1 h from toxin addition was always higher than the 8 : 1 value obtained with Ntx. As the majority of PC and a minority of PE is on the outer layer of the plasma membrane (Verkleij et al. 1973; Fontaine et al. 1979; Shina et al. 1993), such high ratios strongly suggest that the main site of action of these neurotoxins is the outer layer of the plasma membrane. Further evidence in favour of a cell surface hydrolytic action of the SPANs is provided by the results of Fig. 2b, which shows no significant hydrolysis of PS with any of the neurotoxins in CGNs. As almost all PS is confined inside cells, and snake PLA2s readily hydrolyse PS (Verkelij et al. 1973; Napias and Heilbronn 1980; Rosenberg 1997), this finding is in keeping with the above conclusion. The present results do not conflict with previous observations that SPANs enter inside neurons (Pražnikar et al. 2008; Rigoni et al. 2008), but they indicate that the internalized toxins are not very active in phospholipid hydrolysis. However, if their hydrolytic activity is concentrated on defined organelles, even minor local phospholipid hydrolysis may be sufficient to alter dramatically the organelle physiology. This consideration would be particularly relevant for mitochondria, which does interact with SPANs and are affected in their membrane permeability properties (Scorrano et al. 2001; Rigoni et al. 2007).
Limited SPAN hydrolysis of a neuroblastoma cell line
Given the limitation on isolation of pure neuronal cultures outlined in the introduction, we investigated whether transformed neuronal cultures could be used as model systems to probe the lipolytic actions of SPANs. In this respect, we evaluated the actions of SPANs on the NSC34 cell line which was developed by fusing a neuroblastoma with motor neuron-enriched embryonic spinal cord cells and was reported to preserve some characters of motoneurons, the main targets of the SPANs in vivo (Cashman et al. 1992). NSC34 cells differentiated with retinoic acid and exhibited neuronal projections and intercellular contacts. This cell line has been previously used in studies of the biological activity of snake neurotoxins (Petrovic et al. 2004; Pražnikar et al. 2008; Caccin et al. 2009).
Figure 3 shows that Ntx and Tpx caused the release of much lower amounts of LysoPC from differentiated NSC34 cells when compared with CGNs; their activities levelled off after 20 min with a maximum hydrolysis of only 2.5% of the cellular PC content; no increased hydrolysis was found using a high toxin concentration (30 nM). Neuronal bulging is a very sensitive morphological parameter of the in vitro action of SPANs (Rigoni et al. 2004). No bulges were observed after 20 min incubation of differentiated NSC34 incubated with 6 nM Ntx or Tpx, with only a few beginning to appear after 1 h. Taken together, these findings indicate that this neuronal cell line is not a sensitive model to study these neurotoxins. This is presumably because of defective SPAN binding, which is clearly preliminary to hydrolysis. This was demonstrated by incubation of NSC34 cells with the lysoPC + oleci acid mixture, lipids which are released by the PLA2 activity. Athought these cells have a minimal response to incubation with SPANs, these lipid mixture readily causes bulging of nerve terminals identical to that induced by SPANs in primary neurons (Caccin et al. 2009).
Role of calcium-activated cellular PLA2 in the phospholipid hydrolysis of CGNs
A major consequence of SPANs action on neurons is the entry of calcium from the external medium caused by the LysoPL + FA-induced increase in the ion permeability of the plasma membrane (Rigoni et al. 2007). The increased intracellular calcium levels could activate a series of cellular PLA2 that could exert their enzymatic action on internal phospholipids (Orrenius et al. 2003). To test the possibility that at least part of the SPAN-induced phospholipids hydrolysis was because of intracellular PLA2s, we treated CGNs with ionomycin, a well known Ca2+ ionophore. A minimal effect was observed (about 2% of LysoPLs production, data significance estimated with the Student’s t-test, p < 0.05) (Fig. 4). Following pre-incubation with the generic cellular PLA2 inhibitor AA (Vishwanath et al. 1988; Rosenthal et al. 1989; Chandra et al. 2002), no reduction in the LysoPLs production by Tetx was detected, whereas a small decrease (significative with p < 0.05) was observed in the case of Ntx. This latter observation might be because of the slight inhibition of the Ntx enzymatic activity itself by AA as detected by performing an in vitro PLA2 assay (not shown). These findings do not exclude the possibility that PLA2s specific for phospholipids containing in their sn-2 position, a particular FA, such as arachidonic acid or docosohexanoic acid, are activated to produce lipid mediators of inflammation (Piomelli et al. 2007), given that the proportion of these phospholipids with respect to the total amount is very low, and so escape detection.
PLA2 activities of snake pre-synaptic neurotoxins in cultured neurons versus chemical substrates
The PLA2 activity of snake pre-synaptic neurotoxins was always measured in vitro with synthetic substrates which were phospholipids of different chemical nature, usually inserted into liposomes or lipid micelles of different compositions (Rosenberg 1997). These variables and additional factors, including buffer composition and pH values, clearly affected the measured turn-over rates. There are no reports of PLA2 activity of SPANs measured in vivo or in neurons in culture. Such measurement is presented here and the data obtained throw new light on the understanding of action of these neurotoxins. Previously, there existed a large discrepancy between mouse toxicity data (fourth column of Table 2) and their PLA2 activities measured in vitro (third column). Here, we have tested the PLA2 activities of the four neurotoxins used here with a sensitive chemical method to provide a novel and coherent estimation (second column). Indeed, the present assay shows that the highly toxic Tpx, but not Tetx, is also a high turn-over PLA2 (second column). However, here we have additionally obtained PLA2 activity data in cultured neurons (first column), which are more biologically relevant to the in vivo situation. These results (first column) lead to two striking observations: (i) the activities of all the four SPANs in live neurons were much lower than those measured in vitro (compare the first and second columns); (ii) the PLA2 activities of the four neurotoxins in neurons were comparable among each other, as there was only a threefold difference among the least active toxin, Tetx, and the most active one, Ntx. Clearly, the discrepancy between mouse toxicities (last column of Table 2) and PLA2 activities remained, but it was not so unreasonable as it appeared to be before, when the two most toxic toxins (Tpx and Tetx) had the lowest PLA2 activities. The difference remaining after this study might well be accounted for by different pharmacokinetics properties and different binding to the pre-synaptic membrane; this latter aspect has been discussed recently in terms of the quaternary structure of Tpx and Tetx (Montecucco and Rossetto 2008).
Table 2. Comparison between PLA2 activity of the four SPANs on cultured neurons and synthetic substrates and relative toxicity
In vivo PLA2 activity (nmols of PL hydrolysed/ min/nmol of toxin)
In vitro PLA2 activity (nmols of PL hydrolysed/ min/nmol of toxin)
In vitro PLA2 activitya (nmols of PL hydrolysed/ min/nmol of toxin)
Table 2 also shows the large difference existing between the PLA2 activities measured in vitro on synthetic substrates and on neurons in culture. It is very likely that most of this difference is to be attributed to the fact that in the in vitro assay used here the phospholipid substrate of the enzyme is a monomer in solution, endowed with a high accessibility to the active site, whilst the enzyme in neurons acts on a substrate inserted in a membrane which has to be displaced from the lipid bilayer to reach the active site of the enzyme bound to the membrane surface, as discussed before (Reynolds et al. 1991). Moreover, cholesterol was shown to markedly reduce the rate of phospholipid hydrolysis (Stron and Kelly 1977; Napias and Heilbronn 1980).
The present paper reports the first detailed analysis of the PLA2 activity of Ntx, β-Btx, Tpx and Tetx in living neurons which is made possible by modern lipid MS. This analysis revealed novel and unexpected findings that explained apparent contradictions present in the literature. In fact, previously, there appeared to be a strong discrepancy among the PLA2 activities measured in vitro with synthetic substrates and the mouse toxicities of these pre-synaptic snake neurotoxins (Rosenberg 1997; Montecucco and Rossetto 2000). This is summarized in Table 2. Tetx and Tpx were reported to have the lowest in vitro PLA2 activities and yet to be the most toxic of all SPANs. Their PLA2 activity measured here with a very sensitive in vitro assay provided us with higher values with respect to the previous literature, but still difficult to reconcile with their toxicities. The mass spectrometric detection of the hydrolysis of PC, PE and PS catalysed by the four neurotoxins tested here in pure cultured neurons provided PLA2 turn-over values which were comparable among the four SPANs, and expected to be closer to those displayed at the NMJ. These values slightly underestimated PLA2 activity as hydrolysis was only measured for the major phospholipid classes (PC, PE and PS) because other cell phospholipids were present in too low amounts to be considered. However, Tetx and Tpx had PLA2 enzymatic activities on neurons close to those of Ntx and β-Btx, but not higher. Therefore, one must invoke more favourable pharmacokinetics and/or pre-synaptic binding activities to account for the remaining difference in toxicity. Thus, the present work largely resolved a major inconsistency of the previous knowledge on the biological properties of SPANs.
The other major result obtained here, using neurons in culture, was that there was a minimal hydrolysis of PS and that the ratio of LysoPC/LysoPE was close to the PC/PE ratio present on the outer layer of the plasma membrane rather than to that of the entire cell phospholipid composition. This indicates that the major site of phospholipid hydrolysis by all the four SPANs tested here is the pre-synaptic membrane surface. This result should be considered in conjunction with several reports that observed SPANs inside cells (Herkert et al. 2001; Neco et al. 2003; Petrovic et al. 2004; Pražnikar et al. 2008; Rigoni et al. 2008). The present findings indicate that the cytosolic SPAN hydrolysis contributes little, if any, to the overall phospholipid hydrolysis. However, they do not exclude the possibility that limited localized hydrolysis may be of great relevance for intoxication as there are organelles very sensitive to the effect of LysoPLs and FA, such as the mitochondria to which Ntx, β-Btx and Tpx bind (Scorrano et al. 2001; Rigoni et al. 2008).
In summary, the results presented here demonstrate the limited and the site-specific phospholipase activity of this panel of pre-synaptic neurotoxins. The discrepancy between the in vitro PLA2 acitivities of the various SPANs, measured in a vesicular assay, with their activity against intact neuronal cells strongly suggests that the plasma membrane microenvironment of receptor-bound SPAN molecules imposes considerable constraints on their ability to hydrolyse PC. This result may have wide implications for other categories of PLA2 effectors acting on cell membranes.
This work was supported by grants from the Fondazione CARIPARO (Physiopathology of the Synapse), Telethon GGP06133 and Regione Veneto Programma Biotech III.