Insulin Signaling in Bupivacaine-induced Cardiac Toxicity: Sensitization during Recovery and Potentiation by Lipid Emulsion

BUPIVACAINE toxicity is an uncommon but life-threatening event.^1,2  If toxicity occurs in the clinical setting, an infusion of IV lipid emulsion (ILE) is used to accelerate recovery. The mechanism of ILE-based reversal of toxicity includes both a scavenging effect and a direct effect that improves cardiac output.³  This improvement of cardiac output is seen in the absence of toxicity⁴  and contributes to the rapid recovery from toxicity.⁵  However, the cellular signaling underlying this effect is unknown. Recent reports demonstrate that bupivacaine disrupts targets of classical insulin signaling,⁶  including protein kinase B (Akt) and ribosomal protein s6 kinase 1, in cellular models.^7,8  Furthermore, the amide-linked local anesthetics, ropivacaine and lidocaine, disrupt the assembly of phosphoinositide-3-kinase (Pi3k), thereby blocking phosphorylation of Akt,⁹  an effect that is independent of classical sodium channel blockade.¹⁰  Beyond Akt, bupivacaine activates other controllers of glucose homeostasis, including 5′-adenosine monophosphate–activated protein kinase (AMPK),^11,12  an effect that may provide cytoprotection during toxicity.¹³  Conversely, ILE and other fatty acids increase phosphorylation of Akt and other canonical insulinergic targets when used as an adjuvant in recovery from ischemia–reperfusion injury.^14–18  In the absence of toxicity, lipid emulsions modulate AMPK in a number of different tissues.¹⁹  Based on this evidence, we hypothesized that bupivacaine-induced cardiac toxicity adversely affects cellular signaling at targets of glucose homeostasis, including Akt and AMPK,⁶  and that recovery with ILE modifies this process.

Rats were housed as pairs in the Veterinary Medical Unit at the Jesse Brown Veterans Affairs Medical Center (JBVAMC, Chicago, Illinois). Experiments were conducted under sterile conditions in the Veterinary Medical Unit at the JBVAMC. Protocols were approved by the Institutional Animal Care and Utilization Committee of the JBVAMC (IACUC Protocol No. 12-18).

Sprague–Dawley rats (n = 29) weighing between 372 and 426 g were induced with isoflurane in a bell jar before we performed a tracheotomy for intubation and maintained them on 1.2 to 1.75% isoflurane for the remainder of the experiment. Animals were instrumented with a carotid catheter to measure blood pressure, bilateral jugular catheters for infusions, and three electrodes to measure electrocardiogram. After a 30-min equilibration period, animals received 10 mg/kg bupivacaine hydrochloride into the left internal jugular over 20 s to produce a transient asystole. A subset of animals also received adjuvant ILE (10 mg/kg, 30% Intralipid^®; Baxter Pharmaceuticals, USA). At prespecified time points (1.5, 5, and 10 min) after injection, animals were sacrificed by rapid cardiac excision. Tissue was immediately frozen in liquid nitrogen. The peritoneum was opened, and the right kidney was removed and frozen in liquid nitrogen as well. Three animals were pretreated 10 min before bupivacaine with 25 μg/kg of IV Wortmannin (Sigma-Aldrich, USA). In addition, a group of animals (n = 8) received only 10 ml/kg lipid emulsion (over 1 min, with no bupivacaine) and were sacrificed at 3.5 min after infusion (matched to the 5-min bupivacaine + ILE time point). Tissues were prepared in a similar manner. Physiological data were recorded with LabChart 7.0 (ADInstruments, USA).

Sprague–Dawley rats (n = 20) were anesthetized by intraperitoneal injection of 60 mg/kg sodium pentobarbital (Abbott Laboratories, USA). Animals were heparinized before cardiac excision. Hearts were suspended from a Langendorff apparatus (Radnoti & ADInstruments, USA), cannulated at the aortic root, and perfused by roller pump at a rate of 16 ml/min with Krebs–Henseleit buffer (100 mM NaCl, 4.74 mM KCl, 1.18 mM KH2PO4, 1.18 mM MgSO4, 1.00 mM CaCl2, 25.00 mM NaHCO3, 11.50 mM glucose, 4.92 mM pyruvate, and 5.39 mM fumarate), at 37°C and pH 7.40. The buffer was equilibrated with a mixture of oxygen (95%) and carbon dioxide (5%) by passing through a membrane oxygenator. Pressure was transduced from a latex balloon in the left ventricle and recorded using LabChart 7.0. Drugs were delivered by syringe pump to stopcocks approximately 2 cm above the aortic valve. Fourteen hearts were randomized to receive nothing (n = 2), 500 μM bupivacaine over 30 s (n = 6) or 500 μM bupivacaine over 30 s with pretreatment of 50 μM LY294002 (Sigma-Aldrich) over 1 min (n = 6). Two minutes after completion of infusion, hearts were flash-frozen to ensure cardiac bupivacaine concentrations were in accordance with previous experiments.²⁰  The remaining six hearts received 20% ILE (10 mg/kg, 20% Intralipid^®) over 10 min infused to a 1% final concentration. Samples of left ventricle were harvested immediately before infusion and at 10-min and flash-frozen in liquid nitrogen.

All procedures were conducted on ice. The apex of heart and apical pole (cortex only) of the left kidney were isolated. Approximately 100 mg of tissue was dissected, washed, and homogenized in 1.5 ml lysis buffer using an Omni Mixer Homogenizer (Omni International, USA). Lysis buffer comprises 18 mM Tris–HCl, 114 mM NaCl, 0.4% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 9 mM sodium pyrophosphate, 0.9 mM sodium fluoride, 9% glycerol, 10% Triton X-100, 10 mM sodium orthovanadate, 1 mM phenylmethanesulfonyl fluoride, and 1 mM dithiothreitol with pH adjusted to 7.4 and supplemented with Roche Complete Mini EDTA-free tablet (Roche Diagnostic Corporation, USA). Samples were spun at 14K rpm for 10 min to remove insoluble tissue and membranes. Total protein concentration was quantified using a Pierce BCA Protein assay (Thermo Scientific, USA) and iMark microabsorbance plate reader (BioRad, USA). Samples were aliquoted into 50 ml portions and refrozen at −80°C. Twenty micrograms of protein were loaded and run on a 4 to 15% mini-PROTEAN^® TGX gel (BioRad) and then transferred to nitrocellulose membrane (Thermo Scientific, USA). Membranes were blotted for 30 min with 5% bovine serum albumin Cohn fraction V (Sigma-Aldrich) and washed and incubated with rabbit anti-mouse primary antibodies against pT308-Akt, pS473-Akt, total Akt, pT172 AMPK, pS79 acetyl-CoA carboxylase (ACC), pS21/pS9 glycogen synthase kinase-3α/β, pT421 p70 s6 kinase (p70s6k), pS235 ribosomal protein s6, pS1387 tuberous sclerosis 2 (TSC2), pS612 insulin receptor substrate-1 (IRS1), and total glyceraldehyde-3-phosphate dehydrogenase as loading control (all antibodies from Cell Signaling Technologies, USA). Subsequently, membranes were washed and incubated with goat anti-rabbit IgG linked to horseradish peroxidase-peroxidase (Cell Signaling). Luminescence was induced with Amersham ECL Prime (GE Healthcare Bio-sciences, USA) and exposed under darkroom conditions on care-stream autoradiography film (Sigma-Aldrich). Protein quantification was conducted in ImageJ (National Institutes of Health, USA). To control for equal loading, samples were quantified and normalized to total glyceraldehyde-3-phosphate dehydrogenase. For AMPK and Akt, phosphoprotein level was subsequently normalized to total protein level. Because time points were early, we did not anticipate major protein synthesis or degradation, and because no changes were observed in total AMPK and Akt, we did not measure (or normalize) total protein levels for other phosphoproteins.

Glycogen levels were measured with a colorimetric assay (MAK016; Sigma-Aldrich). Left ventricular tissue was isolated, homogenized, and lysed. Lysate was spun and supernatant transferred into tubes for quantification. Glycogen was degraded to glucose with hydrolysis enzyme mix and developed with a colorimetric development enzyme mix. Subsequent coloration coupled to the amount of glycogen in the sample was quantified using an iMark microabsorbance plate reader (BioRad).

Fit curves for bupivacaine concentrations in heart and kidney were developed based on data in Fettiplace et al.³  In brief, concentration of drug in tissue was plotted against time and fit to a single-phase exponential decay with a y-intercept held at zero. As described in the previous publication, recovery of cardiovascular parameters (carotid flow and blood pressure) is coincident with cardiac bupivacaine concentrations dropping below known inhibitory concentrations for voltage-gated ion channels. In those experiments, carotid flow was depressed when cardiac bupivacaine concentrations were greater than 100 nmol/g but flows recovered less than 100 nmol/g and peaked approximately 50 nmol/g. This provided a grouping mechanic based on a “recovery threshold” that reflects recovery of cardiovascular parameters and myocardial drug concentrations either more than or less than approximately 100 nmol/g. This level is consistent with the half-maximal inhibitory concentration (IC₅₀) of bupivacaine for cardiac sodium channels and cardiac calcium channels. Based on data from previous publications,^3,21–23  the IC₅₀ for sodium and calcium channels translate to tissue concentrations of 87 and 80 μM, respectively, for both cardiac and kidney tissue. The greater than 100 μM group includes animals sampled at 1.5 and 5 min and the less than 100 μM group includes animals at the 10-min time point. Animals receiving ILE in the less than 100 μM + ILE group were sampled at the 5-min time point. Kidney was used as a control for contractile status. Recovery was ignored, and groups were analyzed based on kidney bupivacaine concentrations greater than 100 μM.

Omnibus testing was conducted by two-way ANOVA with matched samples grouped by kinase and time point. Post hoc differences were assessed by Sidak tests to control for multiple comparisons. Groupwise α was set at 0.05. Because data were non-normal with large differences in scale and variance, data were ranked before statistical testing to provide normalization.^24,25  Physiological and recovery parameters without grouping components were assessed with the double-sided Mann–Whitney U tests. A pilot set of animals in the bupivacaine only group (control, 1, 5, and 10 min), bupivacaine + ILE group (1, 5, and 10 min), and lipid-only group at 5 min were run to make power calculations. For comparisons against baseline, we determined a sample size of eight (four per group) based on a Cohen d of 2.8, power = 0.8, and α = 0.05 for bupivacaine comparisons. For the bupivacaine + ILE and ILE alone, we expected a robust effect size but anticipated the need for a larger sample size (10) based on a slightly smaller Cohen d (approximately 2.3) due to larger variance of values during recovery. Physiological data and densitometry data are expressed as mean ± SEM.

We used a previously characterized single IV injection model of low-dose bupivacaine toxicity^3,5,26  to assess kinase phosphorylation and dephosphorylation in cardiac tissue in response to a recoverable dose of bupivacaine. The advantage of this model is that after the rapid onset of asystole, animals recover without further intervention (fig. 1A). The gradual improvement in cardiac output is coincident with cardiac bupivacaine content falling below the IC₅₀ for multiple voltage-gated ion channels in the heart (fig. 1B), providing a grouping mechanism based on recovery status and tissue concentration of bupivacaine that reflects the level of channel inhibition (see Materials and Methods: Bupivacaine Concentration Fit Curves and Grouping). Unrecovered (e.g., cardiac bupivacaine greater than 100 μM) animals from the 1.5- and 5-min time points were anticipated to have cardiac bupivacaine concentrations of 301 ± 23 nmol/g (95% CI) and 152 ± 25 nmol/g (95% CI), respectively. Recovered animals (e.g., cardiac bupivacaine less than 100 μM) from the 10-min postchallenge time point were expected to have cardiac bupivacaine concentration of 57 ± 21 nmol/g (95% CI). At the 10-min time point when cardiac bupivacaine was less than 100 μM, rate pressure product (RPP) had improved to 37 ± 2% of baseline (mean ± SEM) compared with 15 ± 6% at the 1.5- and 5-min time points (P = 0.01; fig. 1C).

We examined changes in cellular signaling by Western blotting for phosphoproteins in the insulinergic pathway. We quantified concentrations of Akt phosphorylated at S473 and T308, and the downstream targets glycogen synthase kinase-3β (GSK-3β) at S9, p70s6k at T421, ribosomal protein s6 at S235, and feedback phosphorylation of IRS1 at S612 (fig. 1D). Consistent with previous reports,^7–9  bupivacaine reduced signaling in the Akt pathway with treatment accounting for 48% of the effect (fig. 1E; two-way matched sample ANOVA interaction P = 0.0035, kinase effect P = 0.0035, treatment effect P = 0.0083, matching P < 0.0001). Phosphorylation was decreased on Akt at S473 to 63 ± 5% (mean ± SEM, Sidak post-test P = 0.017), p70s6k at T421 to 50 ± 17% (P = 0.0043), ribosomal protein s6 at S235 to 44 ± 11% (P = 0.0015), and IRS1 at S612 to 31 ± 8% (P = 0.0004) compared with baseline values. However, targets directly downstream of Pi3k (T308 on Akt and S9 on GSK-3β) were less consistently affected with reductions on Akt at T308 to 85 ± 25% of baseline (P = 0.98; 95% CI, 0.23 to 1.33) and GSK-3β at S9 to 75 ± 17% of baseline (P = 0.42; 95% CI, 0.27 to 1.06).

To resolve the discrepancy that targets downstream of the mammalian target of rapamycin complex 1 (mTORC1) including p70s6k and s6 were preferentially affected, we looked to test other signaling targets. Specifically, 5′-adenosine-monophosphate kinase (AMPK) provides a countervailing force at TSC2 that inhibits mTORC1 and downstream targets (fig. 2A) and has been implicated in cellular models of bupivacaine toxicity.^11,12  We probed for phosphorylation of AMPK at T172 and its downstream targets, ACC at S79, and TSC2 at S1387 (fig. 2B). We again found that treatment accounted for the largest portion of the variation (77%) in groupwise comparison (two-way matched sample ANOVA: interaction P < 0.0001, kinase effect P < 0.0001, treatment effect P < 0.0011, matching P < 0.0001). Across all time points, phosphorylation increased to 151 ± 19% of baseline at T172 on AMPK (Sidak: P = 0.0001), 179 ± 32% of baseline at S79 on ACC (P = 0.0001; fig. 2C), and 1,266 ± 258% of baseline at S1387 on TSC2 (P = 0.0001; fig. 2D).

We next analyzed the phosphorylation status (of this pathway) during recovery. We found that at the 10-min time point (when cardiac bupivacaine concentrations were less than 100 μM), there was a marked increase in phosphorylation of key signaling proteins above baseline levels (e.g., hyperphosphorylation). This hyperphosphorylation was observed in Akt at S473 to 378 ± 84% of baseline (Sidak: P < 0.0001) and T308 to 390 ± 64% of baseline (P < 0.0001), GSK-3β at S9 to 293 ± 50% of baseline (P = 0.0005), p70s6k at T421 to 203 ± 29% of baseline (P = 0.0002), and IRS1 at S612 to 186 ± 37% of baseline (P < 0.0001; fig. 3A) in comparison with unrecovered animals (two-way matched subject ANOVA: interaction P < 0.0001, kinase effect P < 0.0001, recovery effect P = 0.002, matching P < 0.0001). We further assessed for biochemical changes by examining the cardiac glycogen content in these hearts and found an increase in glycogen from 2.4 ± 0.17 μM/g to 3.4 ± 0.26 μM/g (Mann–Whitney U test, P = 0.006; fig. 3B). Both these changes are consistent with a sensitization to canonical insulin signaling caused by the loss of negative feedback to IRS1 (fig. 3C).

To confirm that the loss of phosphorylation after bupivacaine challenge was not a contractile or flow-dependent phenomenon, we assayed for phosphorylation changes in the kidney, which experiences very high bupivacaine concentrations in this experimental system (fig. 4A) but is not subject to contractile effects. We assessed phosphorylation of Akt at S473 and T308, as well as phosphorylation of the downstream target GSK-3β at S9 (fig. 4, B and C) at three time points when bupivacaine concentrations were above channel blocking thresholds with an expected concentration of 171 ± 25 nmol/g (95% CI). There was a marked treatment-specific effect (two-way ANOVA, treatment P < 0.0001, 69% of variation) with no interaction (P = 0.93) or kinase (P = 0.93) effect. There was a decrease in Akt phosphorylation at S473 to 56 ± 13% of baseline (n = 8, P = 0.0015, Sidak post-test), a decrease in Akt phosphorylation at T308 to 62 ± 11% of baseline (n = 5, P = 0.0012, Sidak post-test), and a decrease in GSK-3β phosphorylation at S9 to 42 ± 18% of baseline (n = 5, P = 0.0041, Sidak post-test).

We used pharmacological inhibitors to assess whether interfering with this pathway modifies recovery. In the in vivo system, we used the irreversible Pi3k inhibitor—Wortmannin (fig. 5A)—to determine whether sensitization to insulin signaling and subsequent activation of Akt is required for recovery. Animals were pretreated with Wortmannin and 10 min later subjected to the standard bupivacaine infusion, sacrificed at 10 min and phosphorylation of targets downstream from Pi3k were assessed (fig. 5B). Treatment with Wortmannin raised blood pressures by approximately 20 mmHg in one animal but did not have any other prominent effects. Consistent with the effect of Wortmannin, we found that phosphorylation of proteins was reduced in comparison with the animals that were not treated with Wortmannin (fig. 5C; two-way matched sample ANOVA, treatment effect 74% of variation, P = 0.0011, kinase effect P < 0.0001, interaction P = 0.0003, matching P < 0.0001). In addition, Wortmannin blunted recovery from toxicity, reducing cardiac output at 10 min to 12 ± 8% of baseline in contrast to 37 ± 2% of baseline in the animals not treated with Wortmannin (P = 0.03) (fig. 5D). To ensure that this was a cardiac-specific effect, we tested the effect of the specific Pi3K inhibitor LY294002 on bupivacaine toxicity in an isolated heart system with a constant flow state. Animals were acutely treated with LY294002, 1 min before bupivacaine toxicity and then subjected to a bupivacaine challenge. Time to recovery was not different between groups (fig. 5E), and physiological parameters were not different upon recovery (fig. 5F), but pretreatment with LY294002 significantly delayed the time until the occurrence of asystole, with three animals in the LY group not experiencing asystole until after the bupivacaine infusion was stopped (fig. 5G).

Next, we characterized the effect of ILE supplementation on kinase phosphorylation during recovery. Consistent with previous reports,⁵  treatment with ILE accelerated physiological recovery from toxicity (fig. 6A) and within 5 min animals treated with ILE had achieved 98 ± 7% of baseline RPP compared with untreated animals who were at 18 ± 9% of baseline (Mann–Whitney U test, P = 0.002). Although animals in the control group began to recover by 10 min, their cardiac function remained depressed compared with that of the ILE group (37 ± 2% vs. 80 ± 13% compared with baseline RPP for control and ILE, respectively; P = 0.005). As with the untreated group, this recovery was coincident with a time point when cardiac bupivacaine concentrations fell below the IC₅₀ for cardiac sodium channels (fig. 6B). For animals receiving ILE in the less than 100 μM + ILE group, the expected cardiac bupivacaine concentration is 59 ± 23 nmol/g (95% CI).

Next, we assessed phosphorylation of proteins in the insulinergic pathway, including pT308-Akt, pS473-Akt, pS9-GSK-3β, pT421 p70s6k, pS235 ribosomal protein s6, and pS612 IRS1 at predetermined time points (1.5, 5, and 10 min) during recovery after ILE treatment (fig. 7A). Consistent with observations during spontaneous recovery from bupivacaine toxicity, we observed hyperphosphorylation in the pathway at the 10-min time point for all proteins (not pictured). However, after ILE at the 5-min time point when cardiovascular parameters are recovered and cardiac bupivacaine concentration is below the IC₅₀ for cardiac sodium channels (fig. 6, A and B), we found significant differences from baseline values for phosphorylation of key proteins (fig. 7B, two-way ANOVA, treatment effect P = 0.004, kinase effect P < 0.0001, interaction P < 0.0001). Akt was phosphorylated both at S473 (Sidak: P = 0.0026) to 150 ± 23% of baseline and at T308 (P = 0.0004) to 167 ± 10% of baseline. No change in phosphorylation was seen in S9-GSK-3β (98 ± 14% of baseline). Downstream of mTORC1, we observed dephosphorylation of p70s6k and s6 and loss of feedback phosphorylation of IRS1: pT421 p70s6k was reduced to 41 ± 16% (P = 0.0005), S235 ribosomal protein s6 reduced to 52 ± 14% (P = 0.0002), and feedback phosphorylation of S612-IRS1 reduced to 45 ± 12% (P < 0.0001) compared with baseline. We confirmed these effects by regressing the data points across time. For the predominant insulinergic phosphorylation site on Akt (T308), there was a positive linear regression against time as relative phosphorylation levels of T308 fit to a linear slope (R = 0.85) with a non-zero slope of 16 ± 3% min⁻¹ (P < 0.0001; fig. 7C), and a y-intercept of 100 ± 2%. Downstream of mTORC1, relative phosphorylation levels on threonine 421 of p70s6k also fit to a linear slope (R = 0.89) of 28 ± 6% min⁻¹ (P > 0.0001) but with a y-intercept of 55 ± 4% that did not include 100% (fig. 7D).

To assess the source of discordance upstream and downstream of mTORC1, we measured phosphorylation levels of AMPK, ACC (fig. 8A), and TSC2 (fig. 8B), in the presence of adjuvant ILE. Consistent with the untreated condition, treatment with the combination of bupivacaine and ILE leads to a robust treatment-specific effect on phosphorylation (two-way ANOVA, P = 0.0002) but with no kinase or interaction effects. AMPK was phosphorylated at T172 to 145 ± 20% of baseline (P = 0.1125), and TSC2 was phosphorylated at S1387 to 170 ± 11% of baseline (P = 0.0033); notably, the latter value was similar to that in the context of bupivacaine alone (fig. 8C).

Finally, we measured the ability of ILE to change phosphorylation levels of signaling proteins in the absence of toxicity (fig. 9A). For this, we injected animals with ILE (with no accompanying bupivacaine) and sacrificed at a time point matched to the 5-min bupivacaine and ILE time point. At this matched time point, ILE produced a robust treatment effect (two-way ANOVA: P < 0.0001) without kinase or interaction effects. We found that ILE rapidly increased phosphorylation of Akt at both T308 (Sidak: P = 0.0192) and S473 (P = 0.0192; fig. 9B). Furthermore, there was a feedback phosphorylation of IRS1 at S612 (P = 0.0229), a phenomenon that is known to contribute to insulin resistance in peripheral skeletal muscle.^27–29  There were no lipid-induced changes in phosphorylation at downstream targets, including GSK-3β, p70s6k, or s6. We also found that treatment with ILE alone contributed to glycogen accumulation in the heart (fig. 9C). Next, we checked the effect of ILE on AMPK, ACC, and TSC2 phosphorylation to determine whether activation of AMPK resulted from ILE or bupivacaine treatment (fig. 9D). Treatment with ILE in the absence of toxicity had no appreciable effects on phosphorylation of AMPK, ACC, or TSC2 (fig. 9E; two-way matched subjects ANOVA, treatment P = 0.35, kinase P = 0.0978, interaction P = 0.0978).

We found in a rat model that systemic bupivacaine toxicity is a dynamic insult, whereby fundamentally distinct and separate effects on insulinergic signaling (i.e., Pi3K, Akt, IRS1) and glucose homeostasis (i.e., AMPK, GSK-3β) occur during induction of toxicity and recovery from it. Our results agree with previous in vitro studies, wherein lethal or cytotoxic concentrations of bupivacaine induce dephosphorylation of Akt^7–9  and phosphorylation of AMPK.^11,12  Both these kinases integrate signaling at mTORC1 to modulate sensitivity to endogenous insulin³⁰  during recovery from toxicity. As such, we also found that bupivacaine toxicity reduced signaling downstream of Raptor and TSC2 (fig. 10A). The loss of signaling from p70s6k to IRS1 sensitizes insulinergic pathways. Consistent with this, we observed that recovery from bupivacaine-induced cardiac toxicity was associated with hyperactivation of insulinergic targets and an approximately 40% increase in cardiac glycogen stores (fig. 10B), which is associated with better outcomes from cardiac ischemia.³¹  Modulation of glycogen levels by AMPK is asserted to protect against postischemic dysfunction.³²  The sensitization of signaling serves as a protective mechanism to normalize energy processing in settings where metabolism is impaired. Blocking sensitization with Wortmannin, an irreversible inhibitor of Pi3K, interfered with recovery. Transiently blocking Pi3K with the reversible inhibitor—LY20794—provided resistance to toxicity. These findings implicate the importance of this pathway and associated kinases in response to drug toxicity. Wortmannin can also interfere with mTORC1 signaling, so that the importance of the AMPK pathway is reinforced by the findings.

We found that ILE produced rapid changes to insulinergic signaling at Akt and other targets coincident with recovery from toxicity. Treatment with ILE in the absence of bupivacaine toxicity activated insulin signaling at Akt but had no effect on AMPK (fig. 11A). In combination, bupivacaine and ILE provided concurrent activation of both AMPK and Akt. AMPK blocked mTORC1 and downstream targets (p70, s6, and IRS1), while the addition of ILE induced early rephosphorylation of Akt, upstream of mTORC1 (fig. 11B). One of our more interesting findings was the rapidity with which these systems turned on and off. AMPK, Akt, and downstream targets were significantly altered at our earliest time point (1.5 min) and again by 10 min, when Akt was fully rephosphorylated. Lipid emulsion drives rapid physiological changes both in the context of toxicity and in its absence that could be mediated by changes in signaling.^4,33  The involvement of these pathways also points to other treatments for toxicity that could accelerate recovery through modulation of metabolic or signaling pathways.

Local anesthetics are known to modulate energy processing as carnitine deficiency sensitizes animals to bupivacaine toxicity,³⁴  a finding that confirmed the clinical observation that carnitine deficiency predisposes patients to bupivacaine toxicity.^35,36  Furthermore, local anesthetics inhibit mitochondrial carnitine exchange in experimental models,^37,38  and supplementation of ATP can overcome bupivacaine toxicity in myocardial cells.³⁹  Beyond energy production, a number of clinical and experimental observations comport with the conclusion that local anesthetics modify insulin signaling. Hypoglycemic or streptozotocin-induced diabetic animals are more sensitive to bupivacaine cardiac toxicity,^40,41  and insulin provides inotropic support during bupivacaine cardiac toxicity.^42,43  In nontoxic situations, diabetic rats experience extended block durations independent of neurotoxicity^44–47  and patients with poorly controlled diabetes experience extended duration of local anesthetic–induced peripheral nerve blocks.^48–50  Despite the importance of understanding the interaction of local anesthetic effects on diabetic patients,⁵¹  the connection between sensitization to toxicity and glucose handling has previously been incompletely addressed.

Conceivably, improvement in insulin signaling could increase myocardial contractility as both GSK-3β and Akt interact with contractile proteins downstream.^52–55  Multiple experimental models have demonstrated that bupivacaine can reduce infarct size in an ischemia–reperfusion model.^56,57  However, this effect is counterintuitive based on the research demonstrating that bupivacaine and similar amide-linked local anesthetics interfere with antiapoptotic signaling by blocking Akt and downstream targets.^7–9  A sensitization to insulin signaling during recovery would provide a rational explanation for this effect. As an interesting aside, a clinical trial found that ILE unexpectedly reduced blood glucose levels when used as a treatment for xenobiotic drug overdose.⁵⁸  This finding comports with our observed connections between bupivacaine toxicity, lipid, and insulin signaling.

A modification of glucose handling and sensitization of insulinergic signaling provide a new perspective on the mechanisms underlying physiological recovery from bupivacaine toxicity. Our results demonstrated that Wortmannin exacerbated bupivacaine cardiac toxicity. We know from previous studies that high-dose epinephrine, the classic protein kinase A activator, can also interfere with the recovery from bupivacaine toxicity.⁵⁹  Furthermore, local anesthetic toxicity produces an ischemic-like insult because of mitochondrial uncoupling.^37,60  It is conceivable that other drugs that exacerbate ischemia–reperfusion injury (opioid receptor antagonists, toll-like receptor 2 antagonists, K_ATP channel antagonists, bradykinin antagonists, protein kinase C [PKC] inhibitors, etc.) would make bupivacaine toxicity worse. Thus, appropriate controls are needed when studying lipid resuscitation therapy to differentiate worsening of toxicity from inhibition of the beneficial effects of ILE.

From a clinical perspective, insulin has been used in the toxicology community for a number of years as a treatment for drug-induced cardiac toxicity. High-dose insulin provides benefit in experimental models of bupivacaine toxicity^42,43  and is used in emergency rooms for other drug overdoses.⁶¹  In particular, calcium channel blocker (CCB) overdose is associated with a functionally hypoinsulinemic state (due to inhibition of insulin secretion) with secondary hyperglycemia. This allows physicians to follow blood glucose as a measure of both the severity of toxicity⁶²  and the patient recovery.⁶³  Therefore, clinicians can predict recovery from CCB toxicity by observing dropping blood glucose levels, even before cardiovascular recovery. Experimentally, CCBs cause derangements in Pi3K signaling,^64,65  a point in common with bupivacaine. Toxicity from tricyclic antidepressants also causes hyperglycemia,⁶⁶  and treatment is associated with increased insulin sensitivity.⁶⁷  Following on this, high-dose insulin therapy is potentially useful for tricyclic toxicity.^61,68 

Lipid emulsion infusion has complex effects on metabolism, driving insulin resistance via diacylglycerol, PKC, and phosphorylation of IRS1.^27,28  Furthermore, it is known that fatty acids have a rapid uptake phase, which accelerates the cycling of intracellular triacylglycerol stores,⁶⁹  and this driving force is chain length specific and exerts chain length-specific modification on cardiac contractility.⁷⁰  Beyond these basic concepts, the world of lipids in metabolism, signaling, and disease is growing increasingly complex.⁷¹  However, the acute effects of ILE are less well understood. If ILE can effectively modify insulin signaling pathways to counteract detrimental effects of drugs other than local anesthetics, then we may have a better heuristic to decide which drug overdoses are treatable with ILE. In particular, ILE provides benefit in CCB overdose^72–74  and tricyclic overdose,^75,76  but the effect is less clear in β-blocker overdose.⁷⁷  This comports with the aforementioned effects that CCBs and tricyclics have on blood sugar and insulin sensitivity.

Our model lacked cardiac compressions, so that rats experienced hypoperfusion and tissue ischemia before hemodynamic recovery. Given that bupivacaine also produces an ischemia-like insult due to mitochondrial uncoupling, our results might reflect changes in Akt and AMPK secondary to ischemia, ATP depletion, and glycolytic switching and not simply bupivacaine toxicity. If this is the case, then results from studies where ILE is used as an adjuvant for I/R injury^14–17  could be cross-applied to ischemic situations due to local anesthetic toxicity. As noted earlier, this would mean that treatments that adversely impact I/R injury could also worsen recovery from bupivacaine toxicity. In addition, we did not probe all pathways involved in IRS1 sensitization. Both PKC and mitogen-activated protein kinase pathways modulate IRS1 and might do so in the context of bupivacaine toxicity. Furthermore, the inhibitors we used could modify binding of bupivacaine or change entry of drug across the plasma membrane.

Bupivacaine-induced cardiac toxicity activates AMPK and inhibits Akt with integration at TSC2 and loss of feedback from p70s6k to IRS1. This sensitizes cardiac tissue to insulinergic signaling during recovery, leading to hyperactivation of both Akt and GSK-3β and the accumulation of glycogen as cardiac function improves. Preventing hyperactivation with Wortmannin—an irreversible Pi3K inhibitor—exacerbates toxicity and demonstrates the importance of this pathway to recovery. In the absence of toxicity, ILE drives phosphorylation of Akt upstream of mTORC1 but has no effect on the AMPK pathway. During toxicity, ILE also causes an early phosphorylation of Akt without perturbing the activation of AMPK by bupivacaine. As such, AMPK may be the primary actor in regard to toxicity, while Akt may be more involved with recovery. If we use other methods to leverage these protective pathways (i.e., insulinergic sensitization, AMPK activation, and Akt activation), we may be able to optimize treatment for bupivacaine and other drug overdose. Whether similar processes are also operating during other intentional and accidental drug overdose and poisoning merits further research.

Dr. Fettiplace was supported by an American Heart Association (Dallas, Texas) predoctoral fellowship 13PRE16810063 and the Department of Anesthesiology at the University of Illinois Hospital and Health Science Center (Chicago, Illinois). Drs. Weinberg and Rubinstein were funded by a U.S. Veterans Administration (Washington, D.C.) Merit Review and a National Institutes of Health CounterACT grant 1U01NS083457-01.

Dr. Weinberg was awarded a U.S. patent related to lipid resuscitation, is cofounder of ResQ Pharma, Inc. (Chicago, Illinois), with Dr. Rubinstein, and established www.lipidrescue.org, an educational Web site on lipid emulsion as treatment of drug overdose and toxicity. The other authors declare no competing interests.