Investigators in the authors' laboratory previously established the critical participation of the cerulospinal noradrenergic pathway in muscular rigidity elicited by fentanyl. The identification of colocalization of glutamate with tyrosine hydroxylase in most locus ceruleus neurons suggests a role for cerulospinal glutamatergic neurotransmission in fentanyl-induced muscular rigidity. This suggestion and the subtype(s) of glutamate receptors involved were investigated here.
Electromyographic signals activated by bilateral microinjection of 2.5 microg fentanyl into the locus ceruleus were recorded differentially from the left sacrococcygeus dorsi lateralis muscle of adult male Sprague-Dawley rats. The effect of intrathecal administration at the lower lumbar spinal cord of various N-methyl-D-aspartate (NMDA) and non-NMDA receptor antagonists or agonists on this index of muscular rigidity was studied. Rats were under mechanical ventilation, and intravenous infusion of ketamine (30 mg x kg(-1) x h(-1)) was maintained until 10 min before fentanyl was administered.
Microinjection of fentanyl bilaterally into the locus ceruleus increased the root mean square and decreased the mean power frequency values of electromyographic signals. The efficacy of fentanyl to elicit muscular rigidity in this manner was significantly reduced by previous intrathecal administration of either 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), (+)-5-methyl-10,11-dihydro-5H-dibenzo[a,d]cyclohepten-5,10-imine maleate (MK-801), D-(-)-2-amino-5-phosphonovaleric acid (AP5), or (+/-)-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP). Intrathecal administration of kainic acid or NMDA also resulted in significant electromyographic activation.
In addition to the cerulospinal noradrenergic mechanism, the cerulospinal glutamatergic pathway and both NMDA and non-NMDA receptors in the spinal cord may mediate fentanyl-induced muscular rigidity in the rat.
Application of high doses of fentanyl, a synthetic micro-opioid receptor agonist,  is common in cardiac anesthesia.  Along with superior analgesic potency and cardiovascular stability, this narcotic analgesic is also commonly accompanied by muscular rigidity, especially during induction of anesthesia. [1,3] Based on combined physiologic, pharmacologic, histochemical, and immunocytochemical evaluations in the rat, [4–8] our laboratory has established that fentanyl may elicit muscular rigidity by activating spinal motoneurons by acting on the locus ceruleus (LC) in the pons. We further demonstrated the critical participation of the cerulospinal noradrenergic pathway in fentanyl-induced muscular rigidity.
Liu et al.  reported recently that glutamate is colocalized with tyrosine hydroxylase in most LC neurons of the cat, including those neurons that project to the ventral horn of the lumbar spinal cord. They found further in physiologic studies that glutamate is probably involved, at least in part via non-N-methyl-D-aspartate (non-NMDA) receptors, in mediating cerulospinal excitation of lumbar motoneurons. The presence of glutamate-like immunoreactivity is also observed in the LC of rodents. [10,11] Similarly, glutamate-like immunoreactivity  and NMDA, kainate, and alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid binding sites  are present in the dorsal and ventral horn of the spinal cord.
It is therefore possible that, in addition to the cerulospinal noradrenergic mechanism, [4–8] the cerulospinal glutamatergic neurotransmission may also participate in fentanyl-induced muscular rigidity. This hypothesis was evaluated in the present study. We found that the cerulospinal glutamatergic pathway and both NMDA and non-NMDA receptors in the spinal cord mediated muscular rigidity induced by bilateral microinjection of fentanyl into the LC.
The procedures used in this study were approved by the Experimental Animal Committee of the National Yang-Ming University and were in accordance with the National Institutes of Health guide for the care and use of laboratory animals. All efforts were made to minimize animal suffering and reduce the number of animals used.
We used a modification of the procedures described by Wang et al.  to prepare animals for intrathecal drug administration.  Briefly, under pentobarbital sodium anesthesia (50 mg/kg given intraperitoneally), a limited laminectomy was carried out in adult male Sprague-Dawley rats (250–280 g) at the level of the first lumbar vertebrae (L1). A PE-10 catheter (Clay Adams, Becton Dickson, Parsippany, NJ) filled with sterilized saline was inserted tangentially along the dorsal surface of the spinal cord, through an opening of the exposed dura, with its tip lodged at the L4-L5 spinal level. This allowed for application of drugs to the sacral spinal cord, where motoneurons for the sacrococcygeus dorsi lateralis muscle are located.  The catheter was then externalized and secured at the neck region. After intramuscular administration of sodium penicillin (10,000 IU), animals were returned to individual cages in an animal room with the temperature maintained at 24 +/- 0.5 [degree sign] Celsius, and they were allowed free access to standard rat chow and water. Only animals that showed full recovery after 4 or 5 days, with no observable motor deficits, were used in subsequent experiments.
During the experiment, animals were anesthetized initially with ketamine (120 mg/kg given intraperitoneally) so we could do the preparatory surgery. This included tracheal intubation for airway patency and cannulation of the right femoral artery to measure systemic arterial pressure. The arterial catheter (Clay Adams PE-50) was connected to a pressure transducer (Statham P231D; frequency range, DC-200 Hz, Valley View, OH) and in turn to a pressure processor amplifier (Gould G-20–4615–52, Valley View, OH) via which systemic arterial pressure signals were amplified and filtered (frequency range, DC-100 Hz). Both femoral veins were also cannulated for infusion of supplemental ketamine or administration of other drugs. The rectal temperature was maintained at 37 [degree sign] Celsius with a heating pad throughout the experiment.
Electromyographic (EMG) signals were recorded differentially with a pair of platinum needle electrodes (Grass type E2; Quincy, MA) that were inserted into the left sacrococcygeus dorsi lateralis muscle. Bioelectric signals were amplified and filtered (frequency range, 3–1,000 Hz) using a universal amplifier (Gould G-20–4615–58). Electromyographic and systemic arterial pressure signals were displayed on a polygraph (Gould RS3400).
Continuous On-line and Real-time Power Spectral Analysis of Electromyographic Signals
Electromyographic signals were also digitized (Neurocorder DR-484, Neurodata, New York, NY) and stored on videotape. In addition, they were subjected to continuous on-line and real-time power spectral analysis, using a computer algorithm developed in our laboratory  in conjunction with a general-purpose IBM-compatible personal computer. For on-line spectral analysis, electromyographic signals were simultaneously relayed to an analog-to-digital converter (Advantech PCL-818, Shindian, Taiwan) connected to the computer, and they were digitized at a rate of 2,048 Hz.
The digitized signals to be analyzed were truncated into small time segments (or windows). For each time segment, our algorithm first estimated the power density of the spectral components based on fast Fourier transformation. It subsequently quantified the magnitude of electromyographic activity by calculating the root mean square (RMS) value, and evaluated the frequency domain of the electromyographic signals by calculating the mean power frequency (MPF) value of each spectrum. These two values represent, respectively, the number of active motor units and the degree of their synchronization during muscle contraction.  Raw electromyographic signals, their respective two-dimensional spectrogram, together with the values of RMS and MPF calculated for each time segment, were output either graphically or numerically on a monitor, printer, or both. By repeating these procedures continuously, we examined the simultaneous spectral changes of electromyographic signals over time in a real-time and on-line manner.
All recording sessions occurred in a quiet room with minimal physical disturbance. As in our previous study, [8,15] animals were maintained on intravenous infusion of ketamine (30 mg [center dot] kg sup -1 [center dot] h sup -1) until 10 min before fentanyl was administered. They were also mechanically ventilated to maintain end-tidal carbon dioxide to be within 4–5%, as monitored by a capnograph (Datex Normocap, Helsinki, Finland). Intrathecal application of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 65 or 130 nmol), (+)-5-methyl-10,11-dihydro-5H-dibenzo[a,d-]cyclohepten-5, 10-imine maleate (MK-801; 90 or 180 nmol), D-(-)-2-amino-5-phosphonovaleric acid (AP5; 75 or 150 nmol), or (+)-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP; 60 or 120 nmol) was delivered, at a volume of 10 micro liter over 2 min, using a 25-micro liter Hamilton microsyringe. These concentrations were adopted from studies [17,18] in which the chemicals were found to be effective in eliciting their respective stipulated pharmacologic blockade.
In our first series of experiments, the effect of intrathecal administration of each glutamate receptor blocker on baseline electromyographic activity was observed for 20 min. This was followed by bilateral microinjection of 2.5 micro gram fentanyl into the LC, [5,7,8] and then electromyographic activity was monitored for another 20 min. The stereotaxic coordinates for the LC were 2 mm posterior to lambda, 1–1.2 mm from the midline, and 6–6.5 mm below the cortical surface. A volume of 50 nl was delivered on each side to prevent the confounding effect of physical expansion on brain tissues. The order of NMDA or non-NMDA antagonists and the vehicle artificial cerebrospinal fluid (aCSF) was altered randomly to avoid sequential dependency on drug treatments. Thus the aCSF treatment data were collected at different stages of the present study and represent a true vehicle control for all drug treatments.
That the glutamate receptor blockers did not by themselves elicit motor paralysis was evaluated in our second series of experiments. The first series of experiments was repeated, with the exception that animals received microinjection of aCSF into the bilateral LC. Only the effect on electromyographic activity of each blocker, given at its higher intrathecal dose, was studied. In our third series of experiments, we evaluated the action of intrathecal application of kainic acid (0.25 or 2.5 nmol) or NMDA (2 or 20 nmol) on baseline and fentanyl-induced electromyographic activity. The effect of each dose of glutamate receptor agonist, given at a volume of 10 micro liter, on EMG-RMS and EMG-MPF values was observed for 20 min. This was followed by the administration of fentanyl (100 micro gram/kg given intravenously), and electromyographic activity was observed for another 20 min. Animals received continuous intravenous infusions of ketamine (30 mg [center dot] kg sup -1 [center dot] sup -1) in both series of experiments.
Drugs and Chemical Agents
Drugs used in this study included fentanyl citrate (Jannsen, Piscataway, NJ) and ketamine HCl (Parke-Davis, Morris Plains, NJ). The CNQX (HBC complex), MK-801, AP5, CPP, and NMDA were obtained from RBI (Research Biochemical International, Natick, MA), and kainic acid from Sigma Chemical Company (St. Louis, MO). They were prepared with distilled water. One percent Evans blue was mixed with the injection medium to aid histologic verification. To avoid the confounding effect of drug interaction, each animal received only a single dose of each glutamate receptor antagonist or agonist.
The brain stem and spinal cord were removed after each experiment and fixed in 10% formaldehyde-saline in 30% sucrose solution for at least 48 h. Histologic verification of the tip of the microinjection needle or the intrathecal cannula was carried out on frozen 25-micro meter sections stained with Cresyl violet.
Data Analysis and Statistical Evaluation
The digital output of results from computer analysis of electromyographic signals for each 64-s time segment was collected for data analysis. The averaged value of all time segments over the time period stipulated for a particular treatment was used for statistical evaluations. We used two-way analysis of variance with repeated measures to assess the difference between treatment groups. This was followed by the Dunnett or Scheffe multiple-range test for a posteriori comparisons of individual means as appropriate. P < 0.05 was taken to indicate statistical significance.
Effect of Intrathecal Administration of Artificial Cerebrospinal Fluid on Fentanyl-Induced Muscular Rigidity
In control experiments, animals were pretreated with intrathecal aCSF. As we observed previously, [5,7,8] subsequent bilateral microinjection of 2.5 micro gram fentanyl into the LC (see, for example, Figure 2, Figure 3, Figure 4, Figure 5) elicited muscular rigidity. This was manifested as a discernible increase in RMS values of the electromyographic signals within 5 min after application of fentanyl. There was a trend of decrease in MPF values, which became statistically significant 15 min after injection. This suggests that the recruited motor units were mobilized progressively into a synchronous action. 
As we reported previously,  fentanyl delivered to sites adjacent to the LC, such as dorsal tegmental gray, or mesencephalic nucleus of the trigeminal nerve, produced minimal effect on electromyographic activity in the aCSF pretreatment group. Thus only results obtained from animals with histologically verified microinjection sites within the confines of the LC (Figure 1) were used in the statistical analysis.
Effect of Intrathecal Administration of CNQX, MK-801, AP5, or CPP on Fentanyl-Induced Muscular Rigidity
Pretreatment with the intrathecally administered non-NMDA antagonist CNQX (65 or 130 nmol) resulted in significant reversal of the increase in RMS and decrease in MPF values of electromyographic signals (Figure 2) elicited by microinjection of 2.5 micro gram fentanyl into the bilateral LC. Whereas a lower dose (90 nmol) of the noncompetitive NMDA receptor blocker MK-801 was ineffective (Figure 3), pretreatment with a higher dose (180 nmol) resulted in discernible antagonism of fentanyl-induced muscular rigidity (Figure 3). Similarly, application of either competitive NMDA antagonist, AP5 (75 or 150 nmol) or CPP (60 or 120 nmol), into the subarachnoid space of the lumbar spinal cord was efficacious in blunting the electromyographic activation by fentanyl (Figure 4, Figure 5). We noted that none of these antagonists by themselves appreciably affected the baseline electromyographic activity.
Effect of Intrathecal Administration of CNQX, MK-801, AP5, or CPP on Baseline Electromyographic Activity
We were concerned that, instead of specific pharmacologic blockade, the significant antagonism by all four glutamate receptor antagonists of fentanyl-induced muscular rigidity may simply result from motor paralysis. Thus we extended our observation of the effect of intrathecal application of CNQX (130 nmol), MK-801 (180 nmol), AP5 (150 nmol), or CPP (120 nmol) on baseline electromyograms to 40 min in animals that received bilateral microinjection of aCSF to the LC. Figure 6shows that neither intrathecal administration of NMDA or non-NMDA receptor antagonists nor local application of aCSF to the LC elicited significant changes in EMG-RMS or EMG-MPF values.
Effect of Intrathecal Administration of Kainic Acid or NMDA on Baseline and Fentanyl-Induced Electromyographic Activity
Administration of kainic acid (0.25 or 2.5 nmol) into the subarachnoid space of the lumbar cord resulted in a significant increase in EMG-RMS, without discernible changes in EMG-MPF values (Figure 7). Whereas a lower dose (2 nmol) induced indiscernible effects, intrathecal administration of NMDA (20 nmol) also appreciably enhanced EMG activity (Figure 8). These response patterns essentially remained unaltered on subsequent administration of fentanyl (100 micro gram/kg given intravenously), which elicited a significant increase in EMG-RMS and decrease in EMG-MPF values in control animals (Figure 7, Figure 8).
The present study shows that pretreatment with intrathecally administered NMDA and non-NMDA receptor antagonists substantially blunted electromyographic activation evoked by microinjection of fentanyl into the bilateral LC. These results extended our previous conclusion to suggest that, in addition to the cerulospinal noradrenergic mechanism, the cerulospinal glutamatergic neurotransmission and both non-NMDA and NMDA receptors in the spinal cord may also mediate fentanyl-induced muscular rigidity.
Glutamate functions via NMDA and non-NMDA receptors as an excitatory transmitter in mono-and polysynaptic reflexes in the spinal cord.  Other studies  indicate that NMDA and non-NMDA receptors are involved in mediating monosynaptic descending control of spinal motoneurons from the brain stem. Of direct relevance to the present study is the recent anatomic and physiologic report by Liu et al.  They showed that the descending excitatory effect of LC on lumbar motoneurons is at least partially mediated by glutamatergic neurotransmission via non-NMDA receptors. It follows that activation of micro-opioid receptors at the LC by fentanyl may result in the release of glutamate at the spinal cord. The subsequent excitatory action on spinal motoneurons, exerted via NMDA and non-NMDA receptors, was then manifested by the observed increase in electromyographic activity.
The above notion was based on results obtained from three complementary series of experiments. First, electromyographic activation by microinjection of fentanyl bilaterally into the LC was unequivocally antagonized by prior intrathecal administration of either of the two categories of antagonist for excitatory amino acids. For example, CNQX is a potent, competitive kainate/quisqualate (non-NMDA) receptor antagonist.  MK-801 is a highly potent and selective noncompetitive NMDA receptor antagonist that acts on NMDA receptor-operated ion channels as an open channel blocker.  CPP is a potent and competitive antagonist, with high specificity and binding affinity for NMDA receptors.  AP5 is a competitive NMDA antagonist that is more hydrophilic than CPP, allowing it to be effectively localized in the subarachnoid space.  Second, the significant antagonism by all four glutamate receptor blockers of fentanyl-induced muscular rigidity was related to specific pharmacologic blockade rather than to nonspecific motor paralysis. Third, intrathecal application of either kainic acid or NMDA similarly resulted in electromyographic excitation. Further, activation of the non-NMDA or NMDA receptors in the lumbar spinal cord in this manner essentially blocked fentanyl-induced muscular rigidity, possibly via desensitization of these glutamate receptors.
Regardless of the relative role of cerulospinal noradrenergic or glutamatergic systems in fentanyl-induced muscular rigidity, it is imperative that motoneurons in the spinal cord must be excited by the cerulospinal pathway before electromyographic activation by fentanyl is realized. There are at least two plausible, although not mutually exclusive, operational schemes. First, fentanyl-induced muscular rigidity may involve disinhibition of spinal motoneurons via an action of norepinephrine and glutamate on separate neuronal populations in the spinal cord. This operational scheme arises from observations that in brain slice or dissociated cell preparations, [25,26] opioids inhibit LC neurons via the opening of K sup + channels. Activation of micro-opioid receptors by opiate agonists inhibits the discharge of noradrenergic neurons in the LC, [27,28] causing in turn a decrease in the release of norepinephrine from the cerulospinal terminals.  As such, electromyographic activation by fentanyl exerted via the cerulospinal pathway would be expected to disinhibit spinal motoneurons. 
For this scheme to be operable, it would require that cerulospinal noradrenergic afferents terminate on glutamatergic interneurons. In this regard, glutamatergic interneurons are known to be present in the ventral horn  and exert their action on spinal motoneurons through NMDA and non-NMDA receptors.  We recently showed that destruction of the noradrenergic terminals, in the lumbar spinal cord by pretreatment with the neurotoxin, DSP4,  or intrathecal administration of the alpha1-adrenoceptor antagonist prazosin  substantially eliminates electromyographic activation induced by microinjection of fentanyl into the bilateral LC. Similar results were obtained in the present study, when NMDA and non-NMDA antagonists were applied to the spinal subarachnoid space. All these observations suggest that the neuronal populations on which norepinephrine and glutamate act are arranged in series.
Second, fentanyl-induced muscular rigidity may involve direct activation of the cerulospinal pathway and corelease of norepinephrine and glutamate at the spinal cord. In the cat, most of the neurons from the dorsolateral pontine tegmentum, including those from LC, that project to the lumbar spinal cord are immunoreactive to both glutamate and tyrosine hydroxylase.  Coexistence of glutaminase and tyrosine hydroxylase in LC neurons has also been demonstrated in the rat.  Electrical stimulation of the LC augments the activity of the lumbar spinal cord, an action that is blunted by either prazosin or CNQX. [9,34] Thus our observed electromyographic activation induced by microinjection of fentanyl bilaterally into the LC may be the result of a concerted action of both glutamate and norepinephrine in the spinal cord.
Glutamate is proposed as a mediator of the fast excitatory postsynaptic potential in sympathetic preganglionic neurons.  On the other hand, iontophoretic application of norepinephrine generally induces only a slow and small depolarization of spinal motoneurons, although it significantly enhances the excitatory response to glutamate. [36,37] Thus it is of interest that we observed a differential time course of electromyographic response after microinjection of fentanyl into the bilateral LC. Whereas EMG-RMS signals were discernibly enhanced within 5 min after injection, a significant decrease in EMG-MPF was achieved only 15 min after injection. It follows that glutamate released from the activated cerulospinal pathway may be responsible for the rapid recruitment of motor units, which are progressively mobilized into a synchronous action via noradrenergic modulation. Intrathecal administration of kainic acid or NMDA also resulted in an increase in EMG-RMS without discernible changes in EMG-MPF values.
Ketamine is unique in producing an unusual trance-like state known as dissociative anesthesia.  Because ketamine is also a non-NMDA antagonist,  a concern arises that it may affect the results obtained in the present study. Although our data did not allow us to discount this concern, the potential experimental confound of ketamine is considered to be minimal for two reasons. First, as in our previous study,  intravenous infusion of supplementary ketamine was discontinued 10 min before the administration of fentanyl. Under this experimental condition, pretreatments with aCSF or NMDA and non-NMDA antagonists resulted in differential effects on electromyographic activation by bilateral microinjection of fentanyl into the LC. Second, in the presence of intravenous infusion of ketamine, intrathecal administration of kainic acid or NMDA similarly resulted in an increase in electromyographic activity. Further, we observed no additive effect between ketamine and the NMDA or non-NMDA receptor antagonists (Figure 6). Third, ketamine exhibits low binding affinity for NMDA receptors. The Ki for ketamine (4916 +/- 528 nM) is more than 150 times higher than MK-801 (31 +/- 2 nM) in rat brain membranes. 
Multiple brain sites and chemical mediators are known to play a role in opiate rigidity. [4–8,41–43] Results from the present study suggest that, in addition to a cerulospinal noradrenergic mechanism, the cerulospinal glutamatergic and NMDA and non-NMDA receptors in the spinal cord may mediate fentanyl-induced muscular rigidity in the rat.