Volatile Anesthetics Bind Rat Synaptic Snare Proteins

The basic SNARE protein expression constructs used have been described in detail.28Rat syntaxin 1A was expressed from a modified pET-11 vector (pHO2c) to produce syntaxin^1-265with a C-terminal His⁶tag and no transmembrane domain.28Full-length rat SNAP-25B (1-206) with a C-terminal His⁶tag was expressed in a modified pET-11 vector (pHO2d).28The cytoplasmatic domain of rat VAMP 2 (synaptobrevin 2) (196) was subcloned into pET-15b (Novagen, Madison, WI) to produce a protein with an N-terminal His⁶tag.28To make a syntaxin construct with both an N-terminal and C-terminal His⁶tag (H6:STX:H6), the 799-bp Nde  I-Eco  RI fragment from pHO2c was ligated into similarly cut pET-28a. The entire coding sequence and insertion sites were confirmed by dye-terminator sequencing.

All recombinant His⁶-tagged fusion protein con-structs were expressed in BL21(DE3)pLysS or BL21Star®(DE3)pLysS bacteria (Stratagene, La Jolla, CA). Cells were grown in TB (terrific broth) at 37°C to an A⁵⁹⁵of approximately 0.6 OD and induced with 0.5 mm IPTG and harvested after 4 h. Except for SNAP-25B multimer preparations, proteins were purified under nondenaturing conditions by Ni²⁺-NTA affinity chromatography (Qiagen, Valencia, CA) as described in Fasshauer et al. ,28except that a modified lysis buffer was used to decrease nonspecific binding (500 mm NaCl, 20 mm imidazole, 50 mm NaH²PO⁴, 20 mm β-mercaptoethanol, pH 8.0). Imidazole was increased stepwise (50 mm, 200 mm, 400 mm) to elute the proteins. Proteins were dialyzed against fast protein liquid chromatography buffer (FPLC) (20 mm Tris, pH 7.4, 100 mm NaCl, 1 mm EDTA, 1 mm DTT); syntaxin and SNAP-25 were further purified by anion exchange chromatography (Mono-Q-column) on an ÄKTA FPLC (Amersham Bioscience, Piscataway, NJ) using a linearly increasing NaCl gradient in FPLC buffer (100–1,000 mm). Peak fractions were pooled and dialyzed against phosphate-buffered saline (PBS) (1.47 mm KH²PO⁴, 10 mm Na²HPO⁴, 2.7 mm KCl, 137 mm NaCl, pH 7.4) and, if necessary, concentrated by ultrafiltration to a final concentration of 2–4 mg/ml. Protein concentration was measured by absorption at 595 nm by Bradford assay. For SNAP-25B preparations containing mostly multimer, the purification protocol was identical except for the following: β-Mercaptoethanol was not used, and the lysis buffer contained 300 mm NaCl and 10 mm imidazole. The N-terminal His⁶tag on H6:STX:H6 was cleaved for some experiments with thrombin according to manufacturer's protocol (Invitrogen, Carlsbad, CA).

Purified recombinant proteins were characterized by gel electrophoresis, circular dichroism (CD) spectroscopy, and Western blots to ensure adequate purity and correct size and helicity. Gel electrophoresis was performed as described by Fasshauer et al.  28Sodium dodecyl sulfate (SDS)-PAGE was performed with a 5% stacking gel (Tris, pH 6.8) and a 10% resolving gel (Tris, pH 8.8). Except where noted, samples were incubated for 5 min either at room temperature (unboiled) or at greater than 95°C (boiled) before being loaded onto the gel. Nondenaturing PAGE was performed in a similar fashion, except that SDS was not used.

Monomeric SNAP-25B, syntaxin 1A, and VAMP2 were mixed together in an equimolar stoichiometric ratio (1:1:1) in PBS buffer and incubated overnight with rocking at 22°C. This sample was further purified by anion exchange chromatography on a Mono-Q-column. The correct formation of the SNARE complex was confirmed by formation of an SDS- and heat-resistant complex that migrated at the correct size on SDS-PAGE and, upon boiling, separated into the three monomers.

Circular dichroism spectroscopy experiments were performed on a Jasco J-600 spectrometer (Jasco, Inc., Easton, MD) in quartz cuvettes with 0.1 cm path length at 22°C. Samples were diluted in PBS buffer to a final concentration of 5 μm. Spectra (200–250 nm) were an average of five scans.

To confirm the identity of the purified proteins, Western blots were performed. Mouse anti-human monoclonal antibodies against VAMP2 (Chemicon MAB333; Chemicon, Inc., Temecula, CA) and against human syntaxin 1A (Sigma S0664; Sigma-Aldrich, Inc., St. Louis, MO) and goat anti-human polyclonal antisera against SNAP-25B (Santa Cruz SC-7538; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were used and visualized with goat anti-mouse (Sigma A9309 [Sigma-Aldrich, Inc.] for syntaxin; Chemicon AP130P [Chemicon, Inc.] for VAMP2) or rabbit anti-goat HRP-conjugated secondary antibodies according to manufacturers protocols. Specificity was confirmed by lack of immunoreactivity with the other SNARE proteins.

For measurement of binding of VAs to SNARE proteins, the difference between the ¹⁹F-NMR transverse relaxation times (T2) of free and protein-bound VA was used.23,24,27Two hundred thirty-two microliters protein sample in PBS buffer diluted with 88 μl D²O to a final concentration of 20 μm was injected into gastight vials containing VAs at various concentrations. The protein was incubated at 22°C with the anesthetic for 30 min, with vortexing approximately every 5 min. Pilot experiments with binding to bovine serum albumin showed this time was sufficient for full partitioning of anesthetic into the aqueous phase by extraction into heptane and quantification by GC29and maximal binding to protein by NMR. Anesthetic concentrations are expressed as the aqueous phase concentration. After equilibration, the protein sample was transferred via  a gastight syringe into a gastight NMR-tube (Shigemi, Allison Park, PA). The ¹⁹F-NMR relaxation measurements were performed at 500 MHz on a Varian Inova-500 (Varian, Inc., Palo Alto, CA) instrument equipped with Nalorac Bio-Quad probe (Nalorac, Corp., Martinez, CA). The observe channel of this probe was readily tuned to ¹⁹F, where a 90° pulse length of 12.5 μs is routinely achieved. All spin-spin relaxation information was measured at 22°C by the method of Carr, Purcell, Meiboom, and Gill30on the single CF³fluorine resonance present in both halothane and isoflurane. The collection parameters for all relaxation measurements were an interpulse delay time of 200 μs, preparation time of 10 s, and four scans collected at each of 16 echo evolution times ranging from 4 to 4,000 ms, except where noted. Spin-spin relaxation data were analyzed using Bayesian probability theory for curve fitting to estimate decay constants from the spin echoes. All data were well fit to a single exponential decay; in all cases, fitting to a biexponential curve significantly worsened the fit (P < 0.01). Therefore, the anesthetic seems to be in fast exchange between bound and free states. This software (Bayes Analyze) is incorporated into Varian VNMR software (Varian Inc.) and has been developed specifically to extract and statistically analyze NMR parameters (frequencies, amplitudes, decay constants, etc.) from time domain free induction decay data. Curves were compared for statistical differences in decay constants by simultaneous curve fitting and F test using Graphpad Prism software (GraphPad Software, Inc., San Diego, CA). P < 0.01 was the threshold for statistical significance.

Recombinant rat syntaxin 1A (hereafter referred to as syntaxin), SNAP-25B (hereafter referred to as SNAP-25), and VAMP2 (hereafter referred to as VAMP) were made with either an N-terminal or C-terminal His⁶tag (fig. 1A), using pET-based expression constructs.28The proteins were sufficiently pure as judged by PAGE (fig. 1B) and were the correct proteins as judged by their appropriate migration and immunoreactivity on Western blots (figs. 1B and C). By circular dichroism, syntaxin was found to be highly helical and induced helicity in SNAP-25 and VAMP by forming the SNARE complex (fig. 1D), similar to previous reports.28In addition, the SNARE complex was both SDS and heat resistant (see figure 6). Syntaxin and SNAP-25 antibodies labeled at least three SDS-resistant complexes migrating at approximately 55, 120, and 200 kd. Lower and higher order species of these sizes have been typical for SNARE complexes composed of wild-type syntaxin, SNAP-25, and VAMP and are thought to be otherwise identical noncovalently interacting monomeric and multimeric SNARE complexes, respectively.31–33We conclude that our recombinant SNARE proteins and complex are pure and structurally similar to previous reports.

We measured binding of VAs to the SNARE proteins and the SNARE complex by ¹⁹F-NMR relaxation measurements of the CF³moiety of isoflurane and halothane. The spin-spin transverse relaxation time (T2) of smaller molecules such as VAs decreases when bound to larger molecules such as proteins. Using the effects on T2, binding of VAs to model proteins such as bovine serum albumin has been demonstrated.23,24In PBS buffer, the halothane and isoflurane relaxation times were highly reproducible (figs. 2A and B); note that all of the data points from five independent experiments for isoflurane and seven for halothane are shown and generally overlie one another. Recombinant VAMP did not significantly reduce the T2 of either halothane or isoflurane (figs. 2A and B). In addition, a crude host bacterial lysate without the expression vector that was passed over an Ni-NTA column did not reduce T2 (data not shown). Therefore, under these conditions, halothane and isoflurane do not bind with an affinity detectable by NMR to VAMP or bacterial lysate.

Unlike VAMP2, SNAP-25 significantly reduced the T2 of both halothane and isoflurane, indicating binding of a significant fraction of VA to SNAP-25 (figs. 3A and B). We performed a large number of T2 measurements with isoflurane and found a great deal of variability depending on the protein preparation (fig. 3B). With some preparations, T2s were consistently low, indicative of binding; other preparations gave consistently high T2s similar to that measured in buffer. The affinity of SNAP-25 for VAs did not correlate with its ability to form ternary complex or its purity. Upon examination by nondenaturing PAGE, we discovered that SNAP-25 preparations varied in the amount of higher-molecular-weight (MW) species; these higher-MW species migrated as a single band on denaturing PAGE at approximately 26 kd and immunoreacted with SNAP-25 antibody (figs. 1C and 3C). Subsequent ¹⁹F-NMR experiments showed that isoflurane bound to preparations containing SNAP-25 multimers (e.g. , preparation 2) but did not significantly bind to preparations with almost exclusively monomer (e.g. , preparation 8) (figs. 3D and E). The T2 varied with isoflurane concentration from the lowest concentration of isoflurane measured (0.13 mm) to approximately 1 mm and then remained constant up to 5.1 mm (figs. 3D and E). These data are consistent with binding sites on SNAP-25 for isoflurane that saturate around 1 mm. Although the affinity of isoflurane for SNAP-25 cannot be estimated precisely without other methods, a four-parameter logistic fit of the variation in T2 gives an apparent Kd of 330 ± 122 μm, which is similar to the isoflurane EC⁵⁰for anesthesia in humans (human EC⁵⁰= 310 μm).34However, SNAP-25 multimers have not been well documented in vivo ; therefore, the physiologic importance of VA binding to SNAP-25 is unclear.

The T2 values of both halothane and isoflurane were significantly decreased by syntaxin, indicative of binding (figs. 4A and B). However, the binding of either anesthetic was not apparently concentration dependent, at least in a concentration range that starts at the EC⁵⁰for anesthesia for these VAs and extends 15- to 25-fold higher (figs. 4C and D). In addition, halothane and isoflurane did not seem to compete for binding to syntaxin (fig. 4E). The lack of concentration dependence of the T2 can be explained if higher affinity sites with longer bound lifetimes and multiple lower affinity sites with shorter bound lifetimes exist on syntaxin.23For example, using identical NMR techniques coupled with partitioning measurements by gas chromatography, Dubois and Evers23showed bovine serum albumin had both low- and high-affinity binding sites that resulted in the T2 never approaching the T2 of buffer. This property of apparent lack of saturability of the binding sites was explained by the relatively low T2 of bound VA molecules versus  free. For bovine serum albumin, the low- and high-affinity sites were calculated to produce T2s for bound halothane of 6 and 9 ms, respectively; even for the low-affinity sites, this is 140-fold lower than the T2 in buffer. Therefore, even if a small fraction of VA is bound, it will produce a significant reduction in the observed T2 and an apparent lack of saturability. Subsequently, x-ray crystallographic studies have confirmed the presence of multiple halothane binding sites on bovine serum albumin with varying affinities.35Resolution of low- and high-affinity binding sites by observance of concentration dependence of T2 would require NMR measurements at lower VA concentrations. However, the signal/noise ratio was insufficient to measure accurately T2s at [VA] below 100 μm (e.g. , see fig. 3E).

Alternatively, syntaxin could somehow nonspecifically alter the T2s of VAs in a way that does not represent binding, perhaps by changing the physical properties of the solution or by weak interaction of VAs all along the surface of the protein. To test this hypothesis, we set out to make truncated syntaxins that might retain their physical properties but lose VA binding affinity. As a precursor to making these mutant syntaxins, we built an expression construct that produced an otherwise identical recombinant syntaxin but with the addition of a thrombin-cleavable N-terminal His⁶tag along with the existing C-terminal His⁶tag (H6:STX:H6). Surprisingly, we found that this doubly tagged syntaxin did not bind isoflurane (fig. 4F). H6:STX:H6 had a CD spectra and ability to form SNARE complex similar to STX:H6 (fig. 4Gand data not shown). Thrombin cleavage of the N-terminal His⁶tag restored ability of the syntaxin to bind VAs, demonstrating that the protein retained the capacity to bind VAs (fig. 4F). Therefore, despite the apparent lack of saturable or competitive binding sites, VA binding to syntaxin can be blocked without denaturation or changes in secondary structure. This result argues for specific binding sites on the protein, access to which is blocked by the N-terminal tag. Alternatively, more subtle structural changes that do not alter complex formation and are not detectable by CD could obliterate VA binding sites.

The SNARE complex forms a 4-α-helical bundle with a hydrophobic interior that, like syntaxin, is a good candidate to bind VAs. The SNARE complex did significantly reduce the T2 of both halothane and isoflurane, indicative of binding (figs. 5A and B). As for syntaxin, the T2 of isoflurane and halothane in the presence of the SNARE complex did not vary over the concentration ranges tested (figs. 5C and D). Similarly, we could not detect competition between isoflurane and halothane for binding sites on the SNARE complex (fig. 5E). Again, the apparent lack of saturable binding sites can be explained by multiple binding sites with varying affinities. The halothane and isoflurane T2 values were similar in the presence of syntaxin and the SNARE complex. If an equal fraction of VA were bound to the SNARE complex and syntaxin, the observed T2 should be reduced more by the larger molecular species, the SNARE complex. Given that the T2s values are nearly identical, the fraction of VA bound to the SNARE complex must be smaller than that bound to syntaxin. Alternatively, we considered the possibility that VAs bound to contaminating syntaxin rather than ternary complex itself. To examine this hypothesis, we diluted our syntaxin preparation so that the concentration of syntaxin, as assessed by band intensity on a Coomassie-stained gel, was similar to the residual syntaxin monomer in the ternary complex preparations. By serial dilution, we estimated that the residual syntaxin in the ternary complex preparation was at most 2 μm. The isoflurane (0.87 mm) T2 in 2 μm syntaxin was 1.31 ± 0.095 s, which is significantly different than the isoflurane T2 of ternary complex (P < 0.0001, F^1,60= 20). In addition, we formed ternary complex with the doubly His⁶-tagged syntaxin (H6:STX:H6). The ternary complex formed with H6:STX:H6 bound isoflurane, producing a T2 similar to that of complex containing STX:H6 (fig. 5F). Given that H6:STX:H6 and the other components forming the ternary complex do not bind VAs (monomeric SNAP-25 preparations were used for all ternary complex preparations), the VA binding species must be the ternary complex and not contaminating SNARE component proteins.

As shown in figure 1, syntaxin, SNAP-25, and VAMP form lower- and higher-MW SNARE complexes, which are thought to represent noncovalent monomeric and multimeric SNARE complexes.31–33We asked whether the monomeric and/or multimeric SNARE complex species bound isoflurane. After FPLC purification of primarily monomeric or multimeric SNARE complexes (fig. 5G), isoflurane T2 measurements were performed with the two complexes (fig. 5H). Both the primarily low-MW complex and the higher-MW complex significantly reduced the isoflurane T2 compared with buffer, and the higher-MW complex had a significantly lower T2 than the lower-MW complex, as is expected because of the presumed slower diffusion of the heavier species. Therefore, we conclude that the SNARE complex whether as a monomer or a multimer binds isoflurane.

What are the consequences of VA binding to the ternary complex? One attractive possibility that could explain VA inhibition of transmitter release is that VA binding destabilizes the complex. Given that denaturation of the complex in vitro  requires boiling, disassembly of the complex by VAs is unlikely at room temperature; rather, instability of the complex might manifest as an apparent slowing of complex formation, a reduction in the melting temperature of the complex, and/or a reduction in helicity. However, the rate and extent of complex formation was not grossly altered by isoflurane (fig. 6A). The thermal stability of the complex was also not detectably reduced by isoflurane (fig. 6B). Similarly, circular dichroism measurements detected no additional reduction in helicity by isoflurane with increasing temperature. Rather, at intermediate temperatures, isoflurane seemed to increase helicity (fig. 6C). CD measurements at more physiologic temperatures confirmed that 2.9 mm but not 0.58 mm isoflurane slightly but significantly increased the helicity of the SNARE complex (fig. 6D).

Firefly luciferase was the first protein shown to bind VAs.36Although clearly not a clinically relevant VA target, the unequivocal demonstration that VAs can bind to a protein was an important discovery accelerating a paradigm shift in theories of anesthesia away from membrane targets to protein. In the ensuing 20 yr, multiple additional proteins have been shown to bind VAs.23,37,38Only a handful of these might reasonably mediate the clinical effects of VAs, and no proteins localized to the synapse are among them.38In particular, a good presynaptic VA candidate target has been elusive. With the demonstration of shortening of isoflurane and halothane T2 times, indicative of VA immobilization, by a subset of SNARE proteins and the SNARE complex, this work reports the first direct evidence for presynaptic volatile anesthetic binding proteins.

In considering whether SNARE proteins might be relevant VA binding proteins, a few criteria are fundamental. First, relevant targets should bind VAs at concentrations in the range that produces general anesthesia. The aqueous EC⁵⁰s for anesthesia in rat are 290 μm for halothane and 350 μm for isoflurane. SNAP-25, syntaxin, and the SNARE complex detectably bound halothane and isoflurane at the lowest concentrations measured, each below 230 μm. Therefore, the SNARE VA-binding proteins bind halothane and isoflurane at relevant concentrations. Second, a relevant VA target should function in the nervous system, specifically in synaptic transmission. Certainly, syntaxin and the SNARE complex meet this criterion. SNARE proteins and the SNARE complex are mediators of evoked neurotransmitter release at most, if not all, chemical synapses. However, the physiologic relevance of SNAP-25 multimers is unclear.

A third criterion for the relevance of a VA target is that the binding should alter the function of the target in a way that might produce anesthesia. The CD spectrum of the SNARE complex was altered by supraclinical concentrations of isoflurane, indicative of a gross secondary structural alteration by VAs. Clinical concentrations might also produce structural changes not detectable by CD, which measures only the overall helical content of the protein. Alternatively, clinical concentrations of VAs may bind to existing pockets in the SNARE complex and produce little structural change, as has been shown for binding to bovine serum albumine.35If binding of VAs to the SNARE complex has a physiologic effect in vivo , a reasonable mechanism might be that VA binding to the SNARE complex alters calcium responsiveness of SNARE complex-mediated fusion or somehow alters the ability of the complex to catalyze opening of a membrane fusion pore.

A conundrum posed by our results is the reconciliation of synapse selective VA action with a putative VA target present in most, if not all, chemical synaptic terminals. VAs have been shown to inhibit neurotransmitter release from a variety of excitatory synapses across multiple phyla.5–12,39However, release from γ-aminobutyric acid-mediated (GABAergic) and peptidergic terminals is relatively resistant to VAs.40,41Further, potassium-evoked transmitter release, which is a more direct means of producing SNARE complex-mediated fusion, is relatively resistant to VAs.42Therefore, if VA binding to SNARE proteins is indeed relevant to inhibition of transmitter release, the effect of the VA binding or the binding itself must depend on other synapse-specific factors.

Two particularly intriguing synapse-specific proteins are mUNC-13 and SNAP-25. mUNC-13 is a highly conserved syntaxin-binding protein that promotes neurotransmitter release. Different isoforms of mUNC-13 are present at GABAergic and glutamatergic synapses in vertebrates.43–45Our genetic data in C. elegans  show that the N-terminus of syntaxin, which has known interactions with only a few synaptic proteins, one of which is UNC-13, can dominantly antagonize VA action on transmitter release.12The findings reported herein coupled with the genetic results suggest a model where binding and/or the effect of binding to SNARE proteins/complex depend on mUNC-13. This model predicts that in C. elegans , the truncated syntaxin acts by binding to UNC-13 and by an unknown mechanism antagonizes VA binding to t-SNAREs and/or the SNARE complex. Another difference between GABAergic and glutamatergic transmission, at least in rat hippocampal neurons, seems to be the presence of SNAP-25.46Verderio et al.  46were unable to detect either known SNAP-25 isoform in rat GABAergic hippocampal neurons. Further, they showed that the normal differential calcium responses of glutamatergic and GABAergic neurons were dependent on the presence or absence of SNAP-25. These surprising results remain to be confirmed in other species and brain regions; nevertheless, these data offer the possibility that neuronal subtype specific expression of SNAP-25 might explain how VAs can selectively inhibit neurotransmitter release depending on the levels or role of SNAP-25 in those neurons.

A final criterion for the relevance of a VA binding protein is that ablation of the target, either pharmacologically or genetically, results in alterations in sensitivity to VAs. No vertebrate organisms have been reported to be highly resistant to clinical concentrations of VAs. To our knowledge, the only organism reported to be highly VA resistant is the C. elegans  mutant unc-64(md130) . As described in the introduction, this syntaxin mutation produces a truncated syntaxin that dominantly antagonizes VA action. These genetic data were in fact the motivation for examining binding of VAs to SNARE proteins. However, it is not clear how a truncated syntaxin could antagonize VA binding or the effect of binding if the VA target is syntaxin, SNAP-25, or the SNARE complex. At least in the case of syntaxin, binding is clearly sensitive to relatively small structural changes as shown by loss of binding with the doubly His⁶-tagged protein. If SNARE proteins are relevant VA targets, we can explain our genetic results in C. elegans  with a model where truncated syntaxin alters posttranslational modification of or interaction with an accessory protein in a way that reduces the affinity of the SNARE protein/complex for VAs. A direct test of the requirement for SNARE proteins in VA action by mutation is complicated by their essential role in transmitter release and normal behavior. Future work will be aimed at identifying SNARE proteins that normally mediate neurotransmitter release but do not bind VAs to genetically test the relevance of SNARE proteins as VA targets.

Besides the relevance of the protein itself, the relevance of the structural motifs, to which VAs are binding in the SNARE proteins, warrants consideration. Of course, at this time, the VA binding sites on the SNARE proteins have not been defined. However, for syntaxin and the SNARE complex, the interior of the 4-α-helical bundles that they form is a good candidate. Synthetic 4-α-helical bundles have been shown to bind VAs with affinities in the clinical range. Based on quenching of the fluorescence of an interiorly located tryptophan, the site of binding was predicted to be within the interior of the bundle.18–20X-ray crystallographic and spectroscopic data indicate that the interior of the 4-α-helical bundle of syntaxin and the SNARE complex is a hydrophobic environment that might bind hydrophobic VAs. The ability of an N-terminal His⁶tag to block binding of VAs to syntaxin argues that the presence of a 4-α-helical bundle is not itself sufficient for VA binding; rather, relatively subtle changes near but not within the α-helical bundles are capable of blocking binding. This argues that the VA binding site is near the N-terminus or that VAs gain access to their binding site from the N-terminus. The increase in helicity by higher concentrations of isoflurane can be explained by binding within the helical bundle and further stabilizing the hydrophobic interactions that form the bundle. Future studies will be aimed at defining the binding sites on the SNARE proteins and using genetic approaches in C. elegans  to determine their relevance to anesthetic behavioral effects.

The authors thank Alex Evers, M.D. (Henry E. Mallinckrodt Professor of Anesthesiology, Washington University School of Medicine, St. Louis, Missouri), and Jason Berilgen (Technician, Washington University, St. Louis, Missouri) for encouragement and initiation of this project. The authors also thank Phyllis Hanson, Ph.D. (Associate Professor of Cell Biology, Washington University School of Medicine), and Mike Nonet, Ph.D. (Associate Professor of Anatomy and Neurobiology, Washington University School of Medicine), for expression constructs and for their advice and assistance throughout the project with protein expression and purification.