In Vivo Activation of Azipropofol Prolongs Anesthesia and Reveals Synaptic Targets*

Background: Azipropofol is a photoactive analog of the general anesthetic propofol. Results: In vivo photolabeling of tadpoles results in covalent ligand binding to neuronal proteins and prolongation of anesthesia. Conclusion: Reconciling time-resolved gel proteomics with behavioral state allows identification of potential anesthetic targets. Significance: In vivo activation of efficacious photolabels provides a novel approach to investigate mechanisms of general anesthesia. General anesthetic photolabels have been instrumental in discovering and confirming protein binding partners and binding sites of these promiscuous ligands. We report the in vivo photoactivation of meta-azipropofol, a potent analog of propofol, in Xenopus laevis tadpoles. Covalent adduction of meta-azipropofol in vivo prolongs the primary pharmacologic effect of general anesthetics in a behavioral phenotype we termed “optoanesthesia.” Coupling this behavior with a tritiated probe, we performed unbiased, time-resolved gel proteomics to identify neuronal targets of meta-azipropofol in vivo. We have identified synaptic binding partners, such as synaptosomal-associated protein 25, as well as voltage-dependent anion channels as potential facilitators of the general anesthetic state. Pairing behavioral phenotypes elicited by the activation of efficacious photolabels in vivo with time-resolved proteomics provides a novel approach to investigate molecular mechanisms of general anesthetics.

General anesthetics bind many proteins with low affinities (M K D values), hindering analyses of ligand-target interactions. Photoactive analogs of clinically used general anesthetics have been developed to aid molecular studies (1)(2)(3)(4)(5). These probes share physicochemical properties with their parent molecules, retain anesthetic activity, and undergo photolysis under long wave ultraviolet light (UVA) (315-400 nm), a feature that limits damage to cellular macromolecules upon irradiation following equilibration with the ligands. With these compounds, anesthetic binding sites have been mapped on integrin lymphocyte function-associated antigen (3,6), Torpedo nicotinic receptors (7,8), ␤-tubulin (9), PKC (10), and GABA A receptors (11,12), among others. Direct identification of anesthetic substrates from complex homogenates has proceeded with a neurosteroid analog (2) and halothane (13), the latter an unaltered general anesthetic containing a carbon-bromine bond broken by shorter UV wavelengths to create reactive carbon-centered radicals (14).
Despite anesthetic efficacy in vivo, the use of these probes has been limited to in vitro preparations. Electrophysiological evidence supports the concept that covalent incorporation of photoactive anesthetics to binding sites can result in prolonged modulation of functional proteins (15). Although alkylphenol anesthetics are thought to act in part through GABA A receptors, genetic studies prove that other "on-pathway" targets exist (16,17). Thus, we tested the feasibility of activating the photoaffinity probe meta-azipropofol (AziPm) 2 in vivo as a tool to identify novel molecular substrates that contribute to alkylphenol general anesthesia. AziPm is an analog of propofol (2,6-diisopropylphenol) that contains an alkyl diazirinyl group in the meta position of the phenol ring (see Fig. 1) (4). In vivo photolabeling of Xenopus laevis tadpoles equilibrated with AziPm results in a previously unreported behavioral phenotype that we call "optoanesthesia." We describe this in conjunction with unbiased, time-resolved gel proteomics employing a tritiated version of the photolabel. and UV band-pass (ϳ250 -375-nm) filters (lamp and filters from Newport, Stratford, CT). Light intensities (measured with an optical power meter (Thorlabs, Newton, NJ)) were 28.1 microwatts/mm 2 and 27.7 microwatts/mm 2 at 350 and 375 nm, respectively. Albino X. laevis tadpoles (stage 45-47) were purchased from Nasco (Fort Atkinson, WI) and housed in supplied pond water for at least 24 h prior to experiments. Animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Pennsylvania.
Tadpole Immobility Studies-Tadpoles were placed in Petri dishes with propofol or AziPm dissolved in pond water. The same physicochemical parameters determine anesthetic passage across the gills and skin as does passage across the blood brain barrier (18,19). In some experiments, after a 30-min equilibration, tadpoles were transferred to fresh water; in others, after equilibration, tadpoles remained on the bench for a sham control or were exposed to UVA (photolabeled in vivo) before transfer to fresh water. Immobility was defined (and scored) as the percentage of tadpoles that did not swim, twitch, or right themselves throughout a 30-s time window preceding every 10-min interval. Tadpole assays are established measures of anesthetic potency (18,19), and reversible immobility is the most commonly used general anesthetic end point. Alternative causes of immobility in our study (e.g. muscular toxicity) were not ruled out, but should have had additional and toxic features that would have been observed (e.g. cardiac muscle dysfunction, etc). The water temperature was 21-22°C for experiments and changed Ͻ0.5°C throughout any experiment.
In Vivo Photolabeling and Isolation of Neuronal Membranes-Tadpoles were incubated for 30 min with 4 M [ 3 H]AziPm and treated Ϯ UVA for 10 min. After transfer to fresh water, tricaine methanesulfonate (500 mg/liter) was added immediately for the zero time point or at 165 min for the emergence time point, and the tadpoles were placed on ice. After decapitation, brains and spinal cords were removed with forceps under a dissecting microscope and placed in ice-cold 0.32 M sucrose, 5 mM Tris, pH 7.4, supplemented with protease inhibitors. Tissue isolation required less than 15 min following each time point; central nervous system (CNS) tissue was homogenized every 3-5 min using a Teflon/glass homogenizer. CNS homogenates were centrifuged at 100,000 ϫ g for 10 min, washed with isolation buffer, and recentrifuged. The pellet was homogenized in 5 mM Tris, pH 7.4, and centrifuged at 100,000 ϫ g for 10 min, washed, and centrifuged again before resuspension in 2 mM Tris, pH 7.4. An aliquot was removed for a protein assay prior to freezing at Ϫ80°C.
In Vitro Photolabeling-Unexposed tadpoles were anesthetized with tricaine methanesulfonate, and neuronal tissue, dissected as above, was homogenized in sucrose buffer, centrifuged at 100,000 ϫ g for 10 min, washed, and recentrifuged. The pellet was suspended in isolation buffer, and the protein concentration was determined and then diluted to 1 mg/ml in a microcentrifuge tube. 4 M [ 3 H]AziPm Ϯ 400 M propofol was added and, after a brief vortex, the tissue was incubated at 21°C in the dark for 10 min. After transfer to a quartz cuvette (path length, 1 mm), the tissue was photolabeled for 10 min using the same light source as above. The homogenates were then centri-fuged at 100,000 ϫ g, homogenized in 5 mM Tris, recentrifuged at 100,000 ϫ g, washed, and stored at Ϫ80°C in 2 mM Tris.
Scintillation Counting of Neuronal Tissue-Dissected CNS tissue from tadpoles treated with 4 M [ 3 H]AziPm Ϯ UVA was placed in 1 ml of ice-cold 2% SDS, 1% Triton X-100, 5 mM Tris, pH 7.4, supplemented with protease inhibitors. Following homogenization, the protein concentration was determined. 5 and 10 l of the homogenates were added to separate vials in scintillation fluid. The disintegrations per minute (dpm) from each vial were normalized to the corresponding protein amount, and the mean of the two values was used for a single experimental measurement.
IEF/SDS-PAGE-After thawing, 100 g of neuronal membrane protein was centrifuged for 15 min at 15,000 ϫ g. Following removal of the supernatant, the pellet was dissolved in 125 l of 7 M urea, 2 M thiourea, 20 mM dithiothreitol, and 0.2% carrier ampholytes. Isoelectric focusing and SDS-PAGE proceeded according to the manufacturer's instructions, with 3-10 nonlinear pH strips (7 cm) and 4 -15% SDS-PAGE. Tissue from ϳ25 tadpoles sufficed for a single gel.
Spot Intensity Quantitation and Liquid Scintillation Counting-Gels were washed with water and fixed overnight in 15% trichloroacetic acid before staining with Coomassie Blue G-250. After destaining, the gels were scanned on a Bio-Rad GS-800 calibrated densitometer with quantitation performed using the accompanying Quantity One software. Background was subtracted with a box drawn between the 50-and 75-kDa molecular mass markers, and mean optical density multiplied by spot area was recorded from contoured spots.
Spots were excised with a 1.5-mm cylindrical hole punch and placed into scintillation vials. 400 l of 30% hydrogen peroxide was added, and the sealed vials were incubated overnight at 65°C to dissolve the polyacrylamide. These were cooled to room temperature before adding scintillation fluid.
Mass Spectrometry Analysis-Trypsin-digested samples were separated on a nanoLC column before online electrospray into a Thermo LTQ linear ion trap. Raw data were acquired with Xcalibur. The National Center for Biotechnology Information (NCBI) online protein database was searched with the term "Xenopus," and the FASTA formatted sequences of the results were downloaded. This downloaded database was searched with SEQUEST for protein identification. Parameters were 1 atomic mass unit of parent ion tolerance, 1 atomic mass unit of fragment ion tolerance, and 1 missed cleavage. The search result files were combined with Scaffold 3 and filtered with the following criteria: X corr scores (ϩ1 ion) 1.7, (ϩ2) 2.3, (ϩ3) 2.8; protein identification confidence 99.9%; peptide identification confidence 95%; two peptide minimum. Spectra were manually inspected to ensure quality and confidence.
Statistics-GraphPad Prism 5 was used for figure preparation and data analysis. Student's t test, one-way ANOVA, two-way ANOVA, and Bonferroni's correction were calculated within the GraphPad software. Significance is expressed as *, p Ͻ 0.05 and **, p Ͻ 0.01.

RESULTS
Optoanesthesia in Xenopus Tadpoles-Albino tadpoles anesthetized with 3 M propofol or 4 M AziPm (approximate EC 99 doses) recovered on similar time scales following transfer to fresh pond water (Fig. 1A, left). The tadpoles are equilibrated with the alkylphenol, and recovery under these conditions is largely a function of drug diffusion back into the water.
In this study, we hypothesized that covalent occupation of ligand binding sites in vivo would result in prolonged anesthetic effects following washout of unadducted compound. The diazirine of AziPm has a peak absorbance of ϳ370 nm, undergoing photolysis to form a reactive carbene, whereas propofol absorbs wavelengths less than 300 nm. Albino tadpoles immobilized with AziPm and exposed to UVA before transfer to fresh water exhibited prolonged immobility not observed with propofol control groups (Fig. 1A, right). Further, a relationship between lamp exposure time and recovery time was evident, suggesting progressive occupancy of functionally relevant sites. No toxicity (premature death) or differences in body mass were observed between tadpoles treated with either alkylphenol anesthetic or alkylphenol anesthetic with UVA following emergence (measured up to 10 days).
Covalent adduction would concentrate ligand into protein sites with infinitely low off-rates, increasing the apparent potency of the molecule. In vivo photolabeling for 10 min after equilibration with a sub-EC 99 AziPm dose markedly increased the population of immobilized tadpoles (Fig. 1B). Further, we hypothesized that retained attachment of AziPm in functionally relevant targets after washout and emergence would manifest as a decrease in the effective concentration of propofol for immobility. Thus, 20 h after emergence, tadpoles treated as above were exposed to 2 or 0.8 M propofol. Animals photolabeled in vivo displayed increased sensitivity (more rapid induction, slower emergence, and induction with a lower dose) relative to controls (Fig. 1C).
Lastly, 4 M AziPm in pond water was photolyzed for a period corresponding to twice the diazirine half-life (i.e. to a final concentration of ϳ1 M (measured by absorption spectroscopy) plus whatever the product(s) of photolysis are). Tadpoles were then placed in this solution, and after 30 min, immobility was not observed, ruling out the possibility that a more potent, "caged" anesthetic with slower washout kinetics was released with light (data not shown). Together, these data suggest prolonged anesthetic influence due to photoadduction of ligand in vivo, a phenomenon we termed optoanesthesia.
[ 3 H]AziPm in Tadpole Neuronal Tissue-Because general anesthetics are assumed to exert their effects through CNS targets, retention of photoactivated AziPm in neural tissue was measured following optoanesthesia. Brains and spinal cords from control tadpoles and those photolabeled in vivo with [ 3 H]AziPm were isolated to quantify radioactivity after recovery in fresh water ( Fig. 2A). Following [ 3 H]AziPm induction, without washout, no difference is seen between groups treated Ϯ UVA. However, 8-fold more radioactivity was noted in the neuronal tissue of photolabeled animals at 165 min, the point of emergence for all tadpoles exposed to 4 M AziPm and 10 min of UVA.
Identification of Photolabeled Proteins-Optoanesthesia indicates that neuronal substrates photolabeled in vivo are rel-  JANUARY 11, 2013 • VOLUME 288 • NUMBER 2 evant targets of AziPm and possibly propofol anesthesia. For protein identifications, neuronal membrane protein from tadpoles equilibrated with [ 3 H]AziPm and photolabeled for 10 min was subjected to IEF/SDS-PAGE. Duplicate gels were stained, and 100 random spots were excised for scintillation counting (Fig. 2B). Mean background radiation (from three spots excised from each gel containing no detectable Coomassie Blue staining in a region through which the SDS-PAGE separated proteins migrated) was 18.0 dpm, and the mean for all 100 spots was 18.8 dpm. Seven spots contained dpm greater than two standard deviations from the background mean (Fig. 2C). No protein spots from control tadpoles incubated with [ 3 H]AziPm but not exposed to the lamp contained counts exceeding this background threshold.

Light-induced Prolongation of Anesthesia
Response to in Vivo Photolabeling-We hypothesized that for tadpoles to regain movement ("emerge"), the cellular compo-nents contributing to mobility must adapt by removing photolabeled proteins whose activity is altered and/or by replacing these photolabeled macromolecules with newly synthesized proteins. To test this, neuronal membranes were isolated 165 min after tadpoles were photolabeled (when all had emerged) as above for IEF/SDS-PAGE, with the previously identified spots from duplicate gels assayed for dpm. The mean from three spots contained counts within 10% of the initial value (12, 22, and 33), whereas decreases of 46, 35, 42, and 28% were noted in spots 4, 6, 41, and 85, respectively. Coomassie Blue intensity was quantified to assess changes in protein expression, and with the exception of spot 4, little variation was observed (Fig. 2D).
Spot dpm was normalized to corresponding Coomassie Blue intensities for the in vivo experiments. We proposed that proteins with decreased radioactivity content coincident with emergence gain additional credibility as functionally important FIGURE 2. A, quantitation of dpm normalized to protein amount in CNS tissue of tadpoles treated with AziPm Ϯ UVA for 10 min. Data are from 3 experiments per treatment (10 tadpoles per experiment). Mean Ϯ S.E. is shown, and data were analyzed by one-way ANOVA with Bonferroni's post hoc test comparing dpm within each time point (p Ͻ 0.01). B, representative Coomassie Blue-stained gel of tadpole neuronal membrane protein separated first by isoelectric focusing on a 3-10 non-linear (NL) pH gradient strip followed by SDS-PAGE. C, mean dpm of spots excised from gels of tissue isolated immediately after in vivo [ 3 H]AziPm Ϯ UVA treatment. dpm values were arranged in ascending order, with measurements from select spots indicated. The dashed line indicates background mean from ϩ UVA gels with the dotted line indicating two standard deviations. D, Coomassie Blue stain intensity quantified from in vivo gel spots. Spot 4 was found to decrease with a two-tailed Student's t test (p Ͻ 0.05). E, the ratio (dpm/intensity) 165 min divided by (dpm/intensity) 0 min shows the change in the fraction of photolabeled protein over the emergence period. S.D. is shown, and a ratio of 1 would indicate no change.
targets. Thus, we calculated the ratio of normalized photolabel incorporation at 165 min to that at the zero time point for each spot (Fig. 2E). A ratio of 1 would indicate that the fraction of adducted protein did not change over the 165 min. We found that the ratios from spots 6, 33, 41, and 85 are significantly less than 1, suggesting potential relevance in emergence from optoanesthesia.
Conserved Specificity and Target Identification-In vitro photolabeling of neuronal homogenates with 4 M [ 3 H]AziPm Ϯ 400 M propofol was performed to investigate the conserved, saturable specificity of protein sites. Protein spots identified as photolabeled in vivo were analyzed, and all contained dpm above background. A significant effect of propofol on normalized dpm was revealed with decreased photolabel incorporation ranging from 2 to 75% in each spot (p Ͻ 0.05, two-way ANOVA; n ϭ 3 and n ϭ 2 for each spot (Ϫ) and (ϩ) propofol, respectively) ( Table 1). A separate gel was run for protein identification. Six spots were unambiguously identified as containing a single protein, whereas two high confidence identifications were possible in the seventh (Table 1).

DISCUSSION
We describe optoanesthesia, a light-induced anesthetic potentiation, and present a method through which novel general anesthetic targets can be discovered. Reconciling proteomic data with a behavioral phenotype provides a powerful means to assign relevance to identified binding partners. The mechanisms of recovery from optoanesthesia are likely to reside in adducted, relevant proteins being targeted for accelerated degradation and/or being replaced by newly synthesized protein. An alternative hypothesis, not tested here, is that the activities of proteins that are not targets of AziPm are altered to compensate for the covalent modification of the alkylphenol binding partners. Although emergence pathways may be distinct from induction pathways (20), emergence must still require an offloading of the anesthetic from induction targets; thus, we view our initial hypothesis, that label intensity should decrease in functionally relevant targets, as reasonable.
Performing photolabeling in live organisms assured that molecular targets were in a functional state, and because the primary effect of the anesthetic (immobility) was prolonged, confirmed that relevant targets were adducted. Further evidence for ligand incorporation to relevant general anesthetic sites was seen by the increased sensitivity to propofol after in vivo photolabeling. Despite label attachment to the identical proteins in vitro, substantial displacement of photoactive ligand by the parent propofol was most evident with VDACs. These mitochondrial porins with multiple phosphorylation states were also the most prominently labeled proteins in vivo. Protection of photolabeling by propofol suggests conserved and specific alkylphenol site(s). An alternative explanation for propofol competition is allosterism, which would require separate specific cavities for AziPm and for propofol, a possibility we view as unlikely.
Although VDACs are highly abundant proteins, this binding is not interpreted as "nonspecific." Specificity can be viewed as specific to a particular physiological outcome or, on the molecular level (and favored here), high occupancy, saturable binding. Other general anesthetics have been shown to bind VDACs in vitro, but functional consequences have yet to be reported. VDACs bind GABA A receptors (21,22), known propofol targets, but knock-outs of VDAC-1 and VDAC-3 do not appear to alter anesthetic sensitivity (22). Knock-out of VDAC-2 in mice is embryonic lethal, and interestingly, our time-resolved approach implicates VDAC-2 over VDAC-1 and VDAC-3 (Fig.  2E). However, this may reflect a general "protective" effect of VDAC-2 from cellular apoptosis (23).
Of the other identified proteins, SNAP-25 and G␤ 4 exhibited decreases in the ratio of photolabeled to unmodified protein at the time of emergence. Published evidence has suggested that anesthetic interactions with SNAP-25 and/or G␤ 4 might contribute to depressed neuronal signaling. For example, SNAP-25, a component of the ternary SNARE complex, binds volatile anesthetics at physiologically relevant concentrations (24), and isoflurane and propofol inhibit neurotransmitter release through interactions with SNAREs or associated proteins (25,26). Mutagenesis in SNARE complex proteins (including SNAP-25) alters organism sensitivity to general anesthetics (27). Mammalian studies suggest that SNAP-25 may be predominantly expressed in excitatory neurons (28,29), and this protein negatively regulates voltage-gated calcium channels independent of its role in exocytosis (30,31). G␤ (as part of G␤␥) can also directly inhibit presynaptic voltage-gated calcium channels (32,33) and binds to SNAP-25 and syntaxin to inhibit neuronal exocytosis (34,35).
The lack of [ 3 H]AziPm displacement by propofol on some proteins can be interpreted in several ways. For instance, anes-  JANUARY 11, 2013 • VOLUME 288 • NUMBER 2 thetic site(s) (e.g. on SNAREs) may not be conserved among ligands; isoflurane and halothane bind to the SNARE complex in a noncompetitive and nonsaturable manner (24). The hydrophobic interior of the coiled-coil complex likely harbors multiple sites of varying affinities, each capable of binding ligands with low occupancy. Alternatively, protein substrates (including G␤) may be specific to AziPm but not propofol (propofol specificity was tested as the photolabel parent). Regardless, these interpretations do not preclude functional involvement in hypnosis or optoanesthesia, as we did not test competition with nontritiated AziPm, but also, little evidence suggests that saturable binding underlies these states. With our approach in a model vertebrate organism, the tadpole, we provide in vivo evidence for the functional involvement of synaptic targets previously suspected only from in vitro or lower organism studies. Proving this involvement will require extensive genetic manipulations. Additionally, we have identified only a subset of targets bound by AziPm in vivo, not including ion channels (such as the GABA A receptor) that are the object of some general anesthetic hypotheses. Resolving low abundance proteins with multiple transmembrane domains is not feasible with IEF/SDS-PAGE (36,37), and thus, a priori we did not anticipate their identification. The development of proteomic approaches complimentary to IEF/SDS-PAGE, and those that are capable of expanding the dynamic range of neuronal protein detection, will further the investigative power of optoanesthesia.

Light-induced Prolongation of Anesthesia
In conclusion, we anticipate the translation of in vivo photolabeling and behavior-paired proteomics to a wider variety of model organisms and photoactive molecules to investigate molecular mechanisms of general anesthetic pharmacology.